Datasets:

---

DanielPFlorian

/

Transformers-Github-Issues

like 0

[ { "id": 1314768611, "node_id": "MDU6TGFiZWwxMzE0NzY4NjEx", "url": "https://api.github.com/repos/huggingface/transformers/labels/wontfix", "name": "wontfix", "color": "ffffff", "default": true, "description": null }, { "id": 2039044877, "node_id": "MDU6TGFiZWwyMDM5MDQ0ODc3", "url": "https://api.github.com/repos/huggingface/transformers/labels/marian", "name": "marian", "color": "30cc95", "default": false, "description": "" } ]

[ { "login": "sshleifer", "id": 6045025, "node_id": "MDQ6VXNlcjYwNDUwMjU=", "avatar_url": "https://avatars.githubusercontent.com/u/6045025?v=4", "gravatar_id": "", "url": "https://api.github.com/users/sshleifer", "html_url": "https://github.com/sshleifer", "followers_url": "https://api.github.com/users/sshleifer/followers", "following_url": "https://api.github.com/users/sshleifer/following{/other_user}", "gists_url": "https://api.github.com/users/sshleifer/gists{/gist_id}", "starred_url": "https://api.github.com/users/sshleifer/starred{/owner}{/repo}", "subscriptions_url": "https://api.github.com/users/sshleifer/subscriptions", "organizations_url": "https://api.github.com/users/sshleifer/orgs", "repos_url": "https://api.github.com/users/sshleifer/repos", "events_url": "https://api.github.com/users/sshleifer/events{/privacy}", "received_events_url": "https://api.github.com/users/sshleifer/received_events", "type": "User", "site_admin": false } ]

[ "> for NORTH_EU, why is the code \"multilingual\"\r\n\r\nCouldn't find a better qualifier and I think there were a lot of languages\r\n\r\n> how will people know if a model requires a target language code?\r\n\r\nHow do they know now? Maybe add it to the model card? or you could detect if `code.includes('_')`?\r\n\r\n> we will still have many cases where model name is not a language code.\r\n\r\nNot that many, right?", "This issue has been automatically marked as stale because it has not had recent activity. It will be closed if no further activity occurs. Thank you for your contributions.\n" ]

[]

[]

[ "# [Codecov](https://codecov.io/gh/huggingface/transformers/pull/5819?src=pr&el=h1) Report\n> Merging [#5819](https://codecov.io/gh/huggingface/transformers/pull/5819?src=pr&el=desc) into [master](https://codecov.io/gh/huggingface/transformers/commit/63761614eb7a5f96e089b77d5469ea31a7177e16&el=desc) will **decrease** coverage by `1.62%`.\n> The diff coverage is `n/a`.\n\n[](https://codecov.io/gh/huggingface/transformers/pull/5819?src=pr&el=tree)\n\n```diff\n@@ Coverage Diff @@\n## master #5819 +/- ##\n==========================================\n- Coverage 78.59% 76.96% -1.63% \n==========================================\n Files 146 146 \n Lines 26047 26047 \n==========================================\n- Hits 20471 20047 -424 \n- Misses 5576 6000 +424 \n```\n\n\n| [Impacted Files](https://codecov.io/gh/huggingface/transformers/pull/5819?src=pr&el=tree) | Coverage Δ | |\n|---|---|---|\n| [src/transformers/data/data\\_collator.py](https://codecov.io/gh/huggingface/transformers/pull/5819/diff?src=pr&el=tree#diff-c3JjL3RyYW5zZm9ybWVycy9kYXRhL2RhdGFfY29sbGF0b3IucHk=) | `19.81% <0.00%> (-79.28%)` | :arrow_down: |\n| [src/transformers/modeling\\_tf\\_openai.py](https://codecov.io/gh/huggingface/transformers/pull/5819/diff?src=pr&el=tree#diff-c3JjL3RyYW5zZm9ybWVycy9tb2RlbGluZ190Zl9vcGVuYWkucHk=) | `20.27% <0.00%> (-74.92%)` | :arrow_down: |\n| [...rc/transformers/data/datasets/language\\_modeling.py](https://codecov.io/gh/huggingface/transformers/pull/5819/diff?src=pr&el=tree#diff-c3JjL3RyYW5zZm9ybWVycy9kYXRhL2RhdGFzZXRzL2xhbmd1YWdlX21vZGVsaW5nLnB5) | `34.69% <0.00%> (-57.15%)` | :arrow_down: |\n| [src/transformers/data/datasets/glue.py](https://codecov.io/gh/huggingface/transformers/pull/5819/diff?src=pr&el=tree#diff-c3JjL3RyYW5zZm9ybWVycy9kYXRhL2RhdGFzZXRzL2dsdWUucHk=) | `50.74% <0.00%> (-35.83%)` | :arrow_down: |\n| [src/transformers/trainer\\_utils.py](https://codecov.io/gh/huggingface/transformers/pull/5819/diff?src=pr&el=tree#diff-c3JjL3RyYW5zZm9ybWVycy90cmFpbmVyX3V0aWxzLnB5) | `60.00% <0.00%> (-25.72%)` | :arrow_down: |\n| [src/transformers/trainer.py](https://codecov.io/gh/huggingface/transformers/pull/5819/diff?src=pr&el=tree#diff-c3JjL3RyYW5zZm9ybWVycy90cmFpbmVyLnB5) | `16.51% <0.00%> (-21.34%)` | :arrow_down: |\n| [src/transformers/data/processors/glue.py](https://codecov.io/gh/huggingface/transformers/pull/5819/diff?src=pr&el=tree#diff-c3JjL3RyYW5zZm9ybWVycy9kYXRhL3Byb2Nlc3NvcnMvZ2x1ZS5weQ==) | `32.00% <0.00%> (-17.10%)` | :arrow_down: |\n| [src/transformers/training\\_args.py](https://codecov.io/gh/huggingface/transformers/pull/5819/diff?src=pr&el=tree#diff-c3JjL3RyYW5zZm9ybWVycy90cmFpbmluZ19hcmdzLnB5) | `66.32% <0.00%> (-11.23%)` | :arrow_down: |\n| [src/transformers/tokenization\\_roberta.py](https://codecov.io/gh/huggingface/transformers/pull/5819/diff?src=pr&el=tree#diff-c3JjL3RyYW5zZm9ybWVycy90b2tlbml6YXRpb25fcm9iZXJ0YS5weQ==) | `95.89% <0.00%> (-2.74%)` | :arrow_down: |\n| [src/transformers/tokenization\\_xlnet.py](https://codecov.io/gh/huggingface/transformers/pull/5819/diff?src=pr&el=tree#diff-c3JjL3RyYW5zZm9ybWVycy90b2tlbml6YXRpb25feGxuZXQucHk=) | `88.28% <0.00%> (-1.81%)` | :arrow_down: |\n| ... and [5 more](https://codecov.io/gh/huggingface/transformers/pull/5819/diff?src=pr&el=tree-more) | |\n\n------\n\n[Continue to review full report at Codecov](https://codecov.io/gh/huggingface/transformers/pull/5819?src=pr&el=continue).\n> **Legend** - [Click here to learn more](https://docs.codecov.io/docs/codecov-delta)\n> `Δ = absolute <relative> (impact)`, `ø = not affected`, `? = missing data`\n> Powered by [Codecov](https://codecov.io/gh/huggingface/transformers/pull/5819?src=pr&el=footer). Last update [6376161...144b667](https://codecov.io/gh/huggingface/transformers/pull/5819?src=pr&el=lastupdated). Read the [comment docs](https://docs.codecov.io/docs/pull-request-comments).\n" ]

[]

[]

[ "# [Codecov](https://codecov.io/gh/huggingface/transformers/pull/5818?src=pr&el=h1) Report\n> Merging [#5818](https://codecov.io/gh/huggingface/transformers/pull/5818?src=pr&el=desc) into [master](https://codecov.io/gh/huggingface/transformers/commit/63761614eb7a5f96e089b77d5469ea31a7177e16&el=desc) will **decrease** coverage by `1.35%`.\n> The diff coverage is `n/a`.\n\n[](https://codecov.io/gh/huggingface/transformers/pull/5818?src=pr&el=tree)\n\n```diff\n@@ Coverage Diff @@\n## master #5818 +/- ##\n==========================================\n- Coverage 78.59% 77.24% -1.36% \n==========================================\n Files 146 146 \n Lines 26047 26047 \n==========================================\n- Hits 20471 20119 -352 \n- Misses 5576 5928 +352 \n```\n\n\n| [Impacted Files](https://codecov.io/gh/huggingface/transformers/pull/5818?src=pr&el=tree) | Coverage Δ | |\n|---|---|---|\n| [src/transformers/modeling\\_tf\\_mobilebert.py](https://codecov.io/gh/huggingface/transformers/pull/5818/diff?src=pr&el=tree#diff-c3JjL3RyYW5zZm9ybWVycy9tb2RlbGluZ190Zl9tb2JpbGViZXJ0LnB5) | `23.38% <0.00%> (-73.39%)` | :arrow_down: |\n| [src/transformers/modeling\\_tf\\_distilbert.py](https://codecov.io/gh/huggingface/transformers/pull/5818/diff?src=pr&el=tree#diff-c3JjL3RyYW5zZm9ybWVycy9tb2RlbGluZ190Zl9kaXN0aWxiZXJ0LnB5) | `64.90% <0.00%> (-33.90%)` | :arrow_down: |\n| [src/transformers/file\\_utils.py](https://codecov.io/gh/huggingface/transformers/pull/5818/diff?src=pr&el=tree#diff-c3JjL3RyYW5zZm9ybWVycy9maWxlX3V0aWxzLnB5) | `81.19% <0.00%> (-0.30%)` | :arrow_down: |\n| [src/transformers/generation\\_tf\\_utils.py](https://codecov.io/gh/huggingface/transformers/pull/5818/diff?src=pr&el=tree#diff-c3JjL3RyYW5zZm9ybWVycy9nZW5lcmF0aW9uX3RmX3V0aWxzLnB5) | `86.21% <0.00%> (+0.50%)` | :arrow_up: |\n| [src/transformers/modeling\\_tf\\_gpt2.py](https://codecov.io/gh/huggingface/transformers/pull/5818/diff?src=pr&el=tree#diff-c3JjL3RyYW5zZm9ybWVycy9tb2RlbGluZ190Zl9ncHQyLnB5) | `95.32% <0.00%> (+31.77%)` | :arrow_up: |\n| [src/transformers/modeling\\_tf\\_roberta.py](https://codecov.io/gh/huggingface/transformers/pull/5818/diff?src=pr&el=tree#diff-c3JjL3RyYW5zZm9ybWVycy9tb2RlbGluZ190Zl9yb2JlcnRhLnB5) | `93.36% <0.00%> (+49.37%)` | :arrow_up: |\n\n------\n\n[Continue to review full report at Codecov](https://codecov.io/gh/huggingface/transformers/pull/5818?src=pr&el=continue).\n> **Legend** - [Click here to learn more](https://docs.codecov.io/docs/codecov-delta)\n> `Δ = absolute <relative> (impact)`, `ø = not affected`, `? = missing data`\n> Powered by [Codecov](https://codecov.io/gh/huggingface/transformers/pull/5818?src=pr&el=footer). Last update [6376161...f3a250e](https://codecov.io/gh/huggingface/transformers/pull/5818?src=pr&el=lastupdated). Read the [comment docs](https://docs.codecov.io/docs/pull-request-comments).\n", "I don't think you touch this file but, cc @nateraw " ]

[]

[]

[ "Hi! Indeed, I can reproduce. This seems to be an edge-case we were not testing for in late v2+ versions. It does run correctly on v2.3.0, as you've shown, and on recent versions (v3+) as well.\r\n\r\nAfter looking a bit deeper into it, it seems to have happened because of the introduction of `BatchEncoding`, in version v2.9.0. It was later patched in v3.0.0 so the blacklisted versions for using a parallelization mechanism (here the dataloader) with `batch_encode_plus` would be the versions between `v2.9.0` and `v3.0.0`. That would be `v2.9.x`, `v2.10.x` and `v2.11.x`.\r\n\r\nHope this helps, and sorry for the inconvenience." ]

[ { "id": 1314768611, "node_id": "MDU6TGFiZWwxMzE0NzY4NjEx", "url": "https://api.github.com/repos/huggingface/transformers/labels/wontfix", "name": "wontfix", "color": "ffffff", "default": true, "description": null } ]

[]

[ "Hi @psureshmagadi17 , you can add additional layers easily, take a loot the source code for `BERTForSequenceClassification`, you can take that code as it is and add the additional layers before the final classifier. ", "> Hi @psureshmagadi17 , you can add additional layers easily, take a loot the source code for `BERTForSequenceClassification`, you can take that code as it is and add the additional layers before the final classifier.\r\n\r\nHi @patil-suraj , thank you for your response. Did you mean that we can just alter the code in main class? If yes, do you have an example? ", "Hi @psureshmagadi17, if your goal is to add layers to a pretrained model only for fine-tuning BERTForSequenceClassification I think the best option is to modify the BertForSequenceClassification Module.\r\n\r\nIf you want to add attention layers, make sure to use the sequence_output of the BertModel Module and not the pooled_output in the forward function, then use a BertPooler layer before the classifier.", "> Hi @psureshmagadi17, if your goal is to add layers to a pretrained model only for fine-tuning BERTForSequenceClassification I think the best option is to modify the BertForSequenceClassification Module.\r\n> \r\n> If you want to add attention layers, make sure to use the sequence_output of the BertModel Module and not the pooled_output in the forward function, then use a BertPooler layer before the classifier.\r\n\r\nHi @nassim-yagoub - thank you for the response! I'm fairly new to this process i.e., modify the network structure. Do you have an example or discussion that I can follow to help me through this process? ", "A small example:\r\n\r\n import torch.nn as nn\r\n from transformers import BertModel\r\n\r\n class CustomBERTModel(nn.Module):\r\n def __init__(self):\r\n super(CustomBERTModel, self).__init__()\r\n self.bert = BertModel.from_pretrained(\"bert-base-uncased\")\r\n # add your additional layers here, for example a dropout layer followed by a linear classification head\r\n self.dropout = nn.Dropout(0.3)\r\n self.out = nn.Linear(768, 2)\r\n \r\n def forward(self, ids, mask, token_type_ids):\r\n sequence_output, pooled_output = self.bert(\r\n ids, \r\n attention_mask=mask,\r\n token_type_ids=token_type_ids\r\n )\r\n\r\n # we apply dropout to the sequence output, tensor has shape (batch_size, sequence_length, 768)\r\n sequence_output = self.dropout(sequence_output)\r\n \r\n # next, we apply the linear layer. The linear layer (which applies a linear transformation)\r\n # takes as input the hidden states of all tokens (so seq_len times a vector of size 768, each corresponding to\r\n # a single token in the input sequence) and outputs 2 numbers (scores, or logits) for every token\r\n # so the logits are of shape (batch_size, sequence_length, 2)\r\n logits = self.out(sequence_output)\r\n\r\n return logits", "> A small example:\r\n> \r\n> ```\r\n> import torch.nn as nn\r\n> from transformers import BertModel\r\n> \r\n> class CustomBERTModel(nn.Module):\r\n> def __init__(self):\r\n> super(CustomBERTModel, self).__init__()\r\n> self.bert = BertModel.from_pretrained(\"bert-base-uncased\")\r\n> # add your additional layers here, for example a dropout layer followed by a linear classification head\r\n> self.dropout = nn.Dropout(0.3)\r\n> self.out = nn.Linear(768, 2)\r\n> \r\n> def forward(self, ids, mask, token_type_ids):\r\n> sequence_output, pooled_output = self.bert(\r\n> ids, \r\n> attention_mask=mask,\r\n> token_type_ids=token_type_ids\r\n> )\r\n> \r\n> # we apply dropout to the sequence output, tensor has shape (batch_size, sequence_length, 768)\r\n> sequence_output = self.dropout(sequence_output)\r\n> \r\n> # next, we apply the linear layer. The linear layer (which applies a linear transformation)\r\n> # takes as input the hidden states of all tokens (so seq_len times a vector of size 768, each corresponding to\r\n> # a single token in the input sequence) and outputs 2 numbers (scores, or logits) for every token\r\n> # so the logits are of shape (batch_size, sequence_length, 2)\r\n> logits = self.out(sequence_output)\r\n> \r\n> return logits\r\n> ```\r\n\r\nThank you, @NielsRogge", "For example if you want to add the same layers used in Bert, you may want to modify the Module this way (with new_layers_config being the same than the original config, except for the number of layers):\r\n\r\n```python\r\n\r\n\r\nclass BertForSequenceClassification(BertPreTrainedModel):\r\n def __init__(self, config, new_layers_config):\r\n super().__init__(config)\r\n self.num_labels = config.num_labels\r\n\r\n self.bert = BertModel(config)\r\n self.new_layers = BertEncoder(new_layers_config)\r\n self.pooler = BertPooler(config)\r\n self.dropout = nn.Dropout(config.hidden_dropout_prob)\r\n self.classifier = nn.Linear(config.hidden_size, config.num_labels)\r\n\r\n self.init_weights()\r\n\r\n def forward(\r\n self,\r\n input_ids=None,\r\n attention_mask=None,\r\n token_type_ids=None,\r\n position_ids=None,\r\n head_mask=None,\r\n inputs_embeds=None,\r\n labels=None,\r\n output_attentions=None,\r\n output_hidden_states=None,\r\n ):\r\n\r\n outputs = self.bert(\r\n input_ids,\r\n attention_mask=attention_mask,\r\n token_type_ids=token_type_ids,\r\n position_ids=position_ids,\r\n head_mask=head_mask,\r\n inputs_embeds=inputs_embeds,\r\n output_attentions=output_attentions,\r\n output_hidden_states=output_hidden_states,\r\n )\r\n\r\n sequence_output = outputs[0]\r\n \r\n new_layers_output = self.new_layers(sequence_output)[0]\r\n \r\n pooled_output = self.pooler(new_layers_output)\r\n\r\n pooled_output = self.dropout(pooled_output)\r\n logits = self.classifier(pooled_output)\r\n\r\n outputs = (logits,) + outputs[2:] # add hidden states and attention if they are here\r\n\r\n if labels is not None:\r\n if self.num_labels == 1:\r\n # We are doing regression\r\n loss_fct = MSELoss()\r\n loss = loss_fct(logits.view(-1), labels.view(-1))\r\n else:\r\n loss_fct = CrossEntropyLoss()\r\n loss = loss_fct(logits.view(-1, self.num_labels), labels.view(-1))\r\n outputs = (loss,) + outputs\r\n\r\n return outputs # (loss), logits, (hidden_states), (attentions)\r\n\r\n```\r\n\r\nWe added a BertEncoder and a BertPooler to the base implementation.\r\nYou can also retreive the hidden_states and attention of the new layers if you want to, I did not do it here.", "Thanks @nassim-yagoub !", "@nassim-yagoub - I had another question : are the weights for BERTForSequenceClassification Model layers frozen by default?", "The weights are not frozen by default when you load them, however you can manually freeze them with `.requires_grad = False`", "> The weights are not frozen by default when you load them, however you can manually freeze them with `.requires_grad = False`\r\n\r\nThank you @nassim-yagoub!", "This issue has been automatically marked as stale because it has not had recent activity. It will be closed if no further activity occurs. Thank you for your contributions.\n", "> A small example:\r\n> \r\n> ```\r\n> import torch.nn as nn\r\n> from transformers import BertModel\r\n> \r\n> class CustomBERTModel(nn.Module):\r\n> def __init__(self):\r\n> super(CustomBERTModel, self).__init__()\r\n> self.bert = BertModel.from_pretrained(\"bert-base-uncased\")\r\n> # add your additional layers here, for example a dropout layer followed by a linear classification head\r\n> self.dropout = nn.Dropout(0.3)\r\n> self.out = nn.Linear(768, 2)\r\n> \r\n> def forward(self, ids, mask, token_type_ids):\r\n> sequence_output, pooled_output = self.bert(\r\n> ids, \r\n> attention_mask=mask,\r\n> token_type_ids=token_type_ids\r\n> )\r\n> \r\n> # we apply dropout to the sequence output, tensor has shape (batch_size, sequence_length, 768)\r\n> sequence_output = self.dropout(sequence_output)\r\n> \r\n> # next, we apply the linear layer. The linear layer (which applies a linear transformation)\r\n> # takes as input the hidden states of all tokens (so seq_len times a vector of size 768, each corresponding to\r\n> # a single token in the input sequence) and outputs 2 numbers (scores, or logits) for every token\r\n> # so the logits are of shape (batch_size, sequence_length, 2)\r\n> logits = self.out(sequence_output)\r\n> \r\n> return logits\r\n> ```\r\n\r\nI wonder is there a way to do this modification while still using the \"trainer\" just like in the example codes of run_ner.py?\r\n\r\nMy second question is about the return values of self.bert, so basically \"sequence_output\" is the context vector for the sentence, but what does pooled_output mean? ", "> A small example:\r\n> \r\n> ```\r\n> import torch.nn as nn\r\n> from transformers import BertModel\r\n> \r\n> class CustomBERTModel(nn.Module):\r\n> def __init__(self):\r\n> super(CustomBERTModel, self).__init__()\r\n> self.bert = BertModel.from_pretrained(\"bert-base-uncased\")\r\n> # add your additional layers here, for example a dropout layer followed by a linear classification head\r\n> self.dropout = nn.Dropout(0.3)\r\n> self.out = nn.Linear(768, 2)\r\n> \r\n> def forward(self, ids, mask, token_type_ids):\r\n> sequence_output, pooled_output = self.bert(\r\n> ids, \r\n> attention_mask=mask,\r\n> token_type_ids=token_type_ids\r\n> )\r\n> \r\n> # we apply dropout to the sequence output, tensor has shape (batch_size, sequence_length, 768)\r\n> sequence_output = self.dropout(sequence_output)\r\n> \r\n> # next, we apply the linear layer. The linear layer (which applies a linear transformation)\r\n> # takes as input the hidden states of all tokens (so seq_len times a vector of size 768, each corresponding to\r\n> # a single token in the input sequence) and outputs 2 numbers (scores, or logits) for every token\r\n> # so the logits are of shape (batch_size, sequence_length, 2)\r\n> logits = self.out(sequence_output)\r\n> \r\n> return logits\r\n> ```\r\n\r\nIf possible it would be great if you can provide the training loop for the same" ]

[ { "id": 1314768611, "node_id": "MDU6TGFiZWwxMzE0NzY4NjEx", "url": "https://api.github.com/repos/huggingface/transformers/labels/wontfix", "name": "wontfix", "color": "ffffff", "default": true, "description": null } ]

[ { "login": "patrickvonplaten", "id": 23423619, "node_id": "MDQ6VXNlcjIzNDIzNjE5", "avatar_url": "https://avatars.githubusercontent.com/u/23423619?v=4", "gravatar_id": "", "url": "https://api.github.com/users/patrickvonplaten", "html_url": "https://github.com/patrickvonplaten", "followers_url": "https://api.github.com/users/patrickvonplaten/followers", "following_url": "https://api.github.com/users/patrickvonplaten/following{/other_user}", "gists_url": "https://api.github.com/users/patrickvonplaten/gists{/gist_id}", "starred_url": "https://api.github.com/users/patrickvonplaten/starred{/owner}{/repo}", "subscriptions_url": "https://api.github.com/users/patrickvonplaten/subscriptions", "organizations_url": "https://api.github.com/users/patrickvonplaten/orgs", "repos_url": "https://api.github.com/users/patrickvonplaten/repos", "events_url": "https://api.github.com/users/patrickvonplaten/events{/privacy}", "received_events_url": "https://api.github.com/users/patrickvonplaten/received_events", "type": "User", "site_admin": false } ]

[ "Hey @AndreasMadsen, \r\n\r\nThanks a lot for the detailed error description! I added this function, so I will try to take a look as soon as possible. I will be off for the next two weeks though (but will not this issue down). If by change @jplu you find some time, feel free to take a look :-) ", "Should be fixed in the PR https://github.com/huggingface/transformers/pull/5468", "Hi @jplu, I'm sorry, but I doubt #5468 will fix the issue. Fundamentally speaking casting to primitives is not a good practice in TensorFlow, as it invalidates the use of `@tf.function` and is generally unnecessary as described above. Casting to primitives is, in my experience, just never the correct solution in TensorFlow.\r\n\r\nI do think #5468 mitigates the issue, which is maybe where the confusion is coming from. This is because, the models will now correctly default to the `config` object when `output_hidden_states=True` **is not** specified as an input. In those cases object property is never cast to a tensor to begin with, therefore the `@tf.function` graph will be statically compiled to always output the `hidden_states`, as intended.\r\n\r\nHowever, the behavior is different when `output_hidden_states=True` is specified as an input, as it will be cast to a Tensor when it becomes part of the `inputs` argument in `call()`. After that, it is not possible to convert back to a primitive, as that invalidates `@tf.function`.\r\n\r\nIf you insist on keeping it as a primitive, the best solution might be to specify it as an aux-input, similar to `training` and `mask` in a `keras.layers.Layer`, as they don't get converted the same way. I'm not familiar enough with the Keras internals to know the details here, and I think it might also be incompatible with `compute_output_shape` etc. \r\n\r\n_BTW, in the keras RNN layers, hidden_state is only specified in the constructor, properly because it can get a bit messy having to specify it in the `inputs`, but I don't see anything fundamentally wrong with specifying it in `inputs`._", "The PR fix your issue at least for the piece of code you are providing, here the result I have:\r\n\r\n```\r\n>>> import transformers\r\n>>> import tensorflow as tf\r\n>>> bert = tf.function(transformers.TFBertForMaskedLM.from_pretrained('bert-base-uncased'))\r\n>>> for i in range(2):\r\n... (_, hidden_state) = bert(tf.constant([[10,11,12]]), output_hidden_states=True)\r\n... print(f'computed {i}')\r\n... \r\ncomputed 0\r\ncomputed 1\r\n>>> \r\n```", "I believe the solution of @AndreasMadsen is correct -- `output_hidden_states` and also `output_attentions` should be passed as named arguments of the `call` method of `TFBertEncoder.call` -- that way they are not converted to Tensors and can be just constants.", "@jplu Sorry for the misunderstanding. The test now works because `output_hidden_states` is now an auxiliary-input, thus is stays a primitive, thus the casting is no longer involved. However, casting to primitives is still not good practice in TensorFlow, so it doesn't take much to break it.\r\n\r\nI read your code more thoroughly, and have the following two failure cases for you.\r\n\r\n**Case 1:**\r\n\r\n```python\r\nbert = tf.function(transformers.TFBertForMaskedLM.from_pretrained('bert-base-uncased'))\r\noutputs = bert(tf.constant([[10,11,12]]), output_attentions=True)\r\nassert len(outputs) == 2\r\n```\r\n\r\nFails with an IndexError. Essentially because casting in a `tf.function` can not be done. Edit: I think it is an inconsistent casting, in some functions the `output_attentions` is a primitive and the \"casting\" works by being a no-op, it other functions the casting can't be done as it is in a `tf.function`.\r\n\r\n**Case 2:**\r\n\r\n```python\r\nbert = tf.function(transformers.TFBertForMaskedLM.from_pretrained('bert-base-uncased'))\r\noutputs = bert(tf.constant([[10,11,12]]), output_hidden_states=tf.constant(True))\r\nassert len(outputs) == 2\r\n```\r\n\r\noutputs only one output (two is exected), because `tf.constant` can not be casted in a `tf.function`.\r\n\r\n---\r\n\r\nHowever, I would like that, instead of working around these issues, you read the documentation on AutoGraph.\r\n\r\nI really think there is a misunderstanding here, about what TensorFlow AutoGraph can do for you and why casting to primitives is really not necessary at all. I would suggest you read https://www.tensorflow.org/guide/function#conditionals and also check out the hidden docs, such as https://github.com/tensorflow/tensorflow/blob/master/tensorflow/python/autograph/g3doc/reference/control_flow.md#effects-of-the-tracing-process which explains it in more details.\r\n\r\nWhat @foxik says is true, but I think depending on the auxiliary-inputs just avoids the misunderstanding. Truly, casting to primitives is just not a good idea.", "Ok, thanks for the hints, I will review that part! I should be able to start working on it from tomorrow or Thursday.", "I have started to work on this today. I basically just removed the casting to primitive function and updated the code accordingly. Nevertheless it doesn't work when booleans are tensors (your Case 2) because AutoGraph do not let you return an output of different size on each branch. A simple example to simply the use case can be:\r\n\r\n```\r\nimport tensorflow as tf\r\n\r\[email protected]\r\ndef output(output1, output2):\r\n first_conditional_output = ()\r\n second_conditional_output = ()\r\n\r\n for i in range(2):\r\n if output1:\r\n first_conditional_output = first_conditional_output + (i,)\r\n\r\n if output2:\r\n second_conditional_output = second_conditional_output + (i,)\r\n\r\n outputs = (0,)\r\n if output1:\r\n outputs = outputs + (first_conditional_output,)\r\n if output2:\r\n outputs = outputs + (second_conditional_output,)\r\n\r\n return outputs\r\n```\r\n\r\nThis piece of code works fine as long as the parameters are primitives, but when using `tf.constant(True)` for at least one of the two makes it fail with:\r\n\r\n```\r\n/opt/anaconda3/envs/transformers/lib/python3.7/site-packages/tensorflow/python/ops/cond_v2.py:633 _make_indexed_slices_indices_types_match\r\n assert len(set(outs_per_branch)) == 1, outs_per_branch\r\n\r\nAssertionError: [1, 0]\r\n```\r\n\r\nI currently struggling on this, and try to find a workaround but if it is not possible to fix this, I think we will just skip this functionality.", "@jplu TensorFlow guys internally use `tf.cond` to handle this case, for example here: https://github.com/tensorflow/tensorflow/blob/2b96f3662bd776e277f86997659e61046b56c315/tensorflow/python/framework/smart_cond.py#L27-L59 . Keras layers like Dropout have another special case for `Variables` here (the `smart_module.smart_cond` is the above method): https://github.com/tensorflow/tensorflow/blob/2b96f3662bd776e277f86997659e61046b56c315/tensorflow/python/keras/utils/tf_utils.py#L42-L65", "I already came across this, but I don't see how it could solve this issue. Each if in this piece of code is automatically translated to a `tf.cond`. The problem is that there is no else and then the branches `true_fn` and `false_fn` will have a different output size from each other. This is not allowed and this is the problem :(\n\nUnless I'm missing something in what you said.\n\nI'm still trying to figure out how to find a workaround anyway.", "@jplu I thought that `tf.cond` can return arbitrary results from the `true_fn` and `false_fn` (contrary to Autograph), but you are right, it is not allowed during graph construction -- my bad.\r\n\r\nPersonally I think it is enough for the `output_hidden_size` to be a Python constant (and fail otherwise). It is in fact not obvious what type should the output have in case computation graph is used and the `output_hidden_size` is undefined Tensor value.\r\n\r\nSo I think you were right with skipping the functionality ;-)", "@jplu Good point, regarding the variable output-signature. I think it is perfectly acceptable to assert for a primitive input, at least for now.\r\n\r\nAlternatively, the solution would be to return an empty tensor when `tf.constant([False])` is used. Such an approach could look like this:\r\n\r\n```python\r\nimport tensorflow as tf\r\n\r\[email protected]\r\ndef output(output1, output2):\r\n first_conditional_output = tf.TensorArray(tf.int64, size=0, dynamic_size=True, clear_after_read=True)\r\n second_conditional_output = tf.TensorArray(tf.int64, size=0, dynamic_size=True, clear_after_read=True)\r\n\r\n for i in range(2):\r\n if output1:\r\n first_conditional_output = first_conditional_output.write(i, i)\r\n if output2:\r\n second_conditional_output = second_conditional_output.write(i, i)\r\n\r\n outputs = (0,)\r\n if isinstance(output1, tf.Tensor) or output1:\r\n outputs = outputs + (first_conditional_output.stack(),)\r\n\r\n if isinstance(output2, tf.Tensor) or output2:\r\n outputs = outputs + (second_conditional_output.stack(),)\r\n\r\n return outputs\r\n```\r\n\r\nedit: PS: there is more documentation here: https://github.com/tensorflow/tensorflow/blob/master/tensorflow/python/autograph/g3doc/reference/index.md", "@AndreasMadsen Thanks!! I basically already came across the same solution (which is I think the only one :smile:), but the issue here is that we will always have three outputs, which is not really what is wanted. I have to talk to the whole team to see if they are ok with this or not, If yes, we will update the code base accordingly otherwise we will skip this.", "I'm trying to figure out if using `tf.boolean_mask` could work.", "Hi @jplu, I will leave it up to you to decide what is \"wanted\". But you should consider the usage pattern when unpacking the output:\r\n\r\n**with always three outputs**:\r\n\r\n```python\r\[email protected]\r\ndef usage(hasOutput1, hasOutput2):\r\n (one, output1, output2) = output(hasOutput1, hasOutput2)\r\n\r\n tf.print(one)\r\n if hasOutput1:\r\n tf.print(output1)\r\n if hasOutput2:\r\n tf.print(output2)\r\n```\r\n\r\n**with always variable outputs**:\r\n\r\n```python\r\[email protected]\r\ndef usage(hasOutput1, hasOutput2):\r\n\r\n output1 = tf.zeros((0,))\r\n output2 = tf.zeros((0,))\r\n if hasOutput1 and hasOutput2:\r\n (one, output1, output2) = output(hasOutput1, hasOutput2)\r\n elif hasOutput1:\r\n (one, output1) = output(hasOutput1, hasOutput2)\r\n elif hasOutput2:\r\n (one, output2) = output(hasOutput1, hasOutput2)\r\n else:\r\n (one, ) = output(hasOutput1, hasOutput2)\r\n\r\n tf.print(one)\r\n if hasOutput1:\r\n tf.print(output1)\r\n if hasOutput2:\r\n tf.print(output2)\r\n```\r\n", "You are totally right @AndreasMadsen!! Now everything should work like expected in the PR at the condition to use primitive booleans. If we all decide to fix the output size to 3 I will open a new PR for this.\r\n\r\nNevertheless, I've just spotted another problem with the usage of `TensorArray`, instead to have a tuple that looks like : `(batch_size, num_tokens, hidden_states_size) * num_layers` we get a tensor that looks like `(num_layers, batch_size, num_tokens, hidden_states_size)` which cause several other issues for later.", "> Nevertheless, I've just spotted another problem with the usage of TensorArray, instead to have a tuple that looks like : (batch_size, num_tokens, hidden_states_size) * num_layers we get a tensor that looks like (num_layers, batch_size, num_tokens, hidden_states_size) which cause several other issues for later.\r\n\r\nTo avoid a breaking change, you could do it as a tuple of empty tensors. For the next major version, I would suggest it become one big tensor. You can swap `transpose`/swap the first and last axis, to make it mostly compatible, with indexing, unpack, etc..", "Exactly!! This should be a good way to go if the decision of having always 3 outputs is taken.", "This issue has been automatically marked as stale because it has not had recent activity. It will be closed if no further activity occurs. Thank you for your contributions.\n" ]

[ { "id": 1314768611, "node_id": "MDU6TGFiZWwxMzE0NzY4NjEx", "url": "https://api.github.com/repos/huggingface/transformers/labels/wontfix", "name": "wontfix", "color": "ffffff", "default": true, "description": null } ]

[]

[ "Hey @nuradilK,\r\n\r\n1) The GPT2 weight initialization is a known warning bug we should remove, but which does not affect model performance, \r\nsee https://github.com/huggingface/transformers/issues/5800 e.g.\r\n\r\n2) It's very hard for us to track down low performance of specific user code (which data to use, data processing, ...). Do you mind posting a question like \"How to reproduce GPT2 PPL results\" on https://discuss.huggingface.co/ since it is not really a bug but more a high level question. ", "This issue has been automatically marked as stale because it has not had recent activity. It will be closed if no further activity occurs. Thank you for your contributions.\n" ]

[ { "id": 1838412367, "node_id": "MDU6TGFiZWwxODM4NDEyMzY3", "url": "https://api.github.com/repos/huggingface/transformers/labels/model%20card", "name": "model card", "color": "92d5f4", "default": false, "description": "Related to pretrained model cards" } ]

[]

[ "# [Codecov](https://codecov.io/gh/huggingface/transformers/pull/5813?src=pr&el=h1) Report\n> Merging [#5813](https://codecov.io/gh/huggingface/transformers/pull/5813?src=pr&el=desc) into [master](https://codecov.io/gh/huggingface/transformers/commit/89a78be51f1c2afd263da66ec76d3297432f0c2a&el=desc) will **not change** coverage.\n> The diff coverage is `n/a`.\n\n[](https://codecov.io/gh/huggingface/transformers/pull/5813?src=pr&el=tree)\n\n```diff\n@@ Coverage Diff @@\n## master #5813 +/- ##\n=======================================\n Coverage 78.19% 78.19% \n=======================================\n Files 146 146 \n Lines 26047 26047 \n=======================================\n Hits 20367 20367 \n Misses 5680 5680 \n```\n\n\n| [Impacted Files](https://codecov.io/gh/huggingface/transformers/pull/5813?src=pr&el=tree) | Coverage Δ | |\n|---|---|---|\n| [src/transformers/generation\\_tf\\_utils.py](https://codecov.io/gh/huggingface/transformers/pull/5813/diff?src=pr&el=tree#diff-c3JjL3RyYW5zZm9ybWVycy9nZW5lcmF0aW9uX3RmX3V0aWxzLnB5) | `85.96% <0.00%> (-0.26%)` | :arrow_down: |\n| [src/transformers/file\\_utils.py](https://codecov.io/gh/huggingface/transformers/pull/5813/diff?src=pr&el=tree#diff-c3JjL3RyYW5zZm9ybWVycy9maWxlX3V0aWxzLnB5) | `81.49% <0.00%> (+0.29%)` | :arrow_up: |\n\n------\n\n[Continue to review full report at Codecov](https://codecov.io/gh/huggingface/transformers/pull/5813?src=pr&el=continue).\n> **Legend** - [Click here to learn more](https://docs.codecov.io/docs/codecov-delta)\n> `Δ = absolute <relative> (impact)`, `ø = not affected`, `? = missing data`\n> Powered by [Codecov](https://codecov.io/gh/huggingface/transformers/pull/5813?src=pr&el=footer). Last update [89a78be...1965bb0](https://codecov.io/gh/huggingface/transformers/pull/5813?src=pr&el=lastupdated). Read the [comment docs](https://docs.codecov.io/docs/pull-request-comments).\n" ]

[ { "id": 1838412367, "node_id": "MDU6TGFiZWwxODM4NDEyMzY3", "url": "https://api.github.com/repos/huggingface/transformers/labels/model%20card", "name": "model card", "color": "92d5f4", "default": false, "description": "Related to pretrained model cards" } ]

[]

[ "# [Codecov](https://codecov.io/gh/huggingface/transformers/pull/5812?src=pr&el=h1) Report\n> Merging [#5812](https://codecov.io/gh/huggingface/transformers/pull/5812?src=pr&el=desc) into [master](https://codecov.io/gh/huggingface/transformers/commit/89a78be51f1c2afd263da66ec76d3297432f0c2a&el=desc) will **increase** coverage by `0.31%`.\n> The diff coverage is `n/a`.\n\n[](https://codecov.io/gh/huggingface/transformers/pull/5812?src=pr&el=tree)\n\n```diff\n@@ Coverage Diff @@\n## master #5812 +/- ##\n==========================================\n+ Coverage 78.19% 78.50% +0.31% \n==========================================\n Files 146 146 \n Lines 26047 26047 \n==========================================\n+ Hits 20367 20448 +81 \n+ Misses 5680 5599 -81 \n```\n\n\n| [Impacted Files](https://codecov.io/gh/huggingface/transformers/pull/5812?src=pr&el=tree) | Coverage Δ | |\n|---|---|---|\n| [src/transformers/modeling\\_tf\\_distilbert.py](https://codecov.io/gh/huggingface/transformers/pull/5812/diff?src=pr&el=tree#diff-c3JjL3RyYW5zZm9ybWVycy9tb2RlbGluZ190Zl9kaXN0aWxiZXJ0LnB5) | `64.90% <0.00%> (-33.90%)` | :arrow_down: |\n| [src/transformers/generation\\_tf\\_utils.py](https://codecov.io/gh/huggingface/transformers/pull/5812/diff?src=pr&el=tree#diff-c3JjL3RyYW5zZm9ybWVycy9nZW5lcmF0aW9uX3RmX3V0aWxzLnB5) | `86.21% <0.00%> (ø)` | |\n| [src/transformers/modeling\\_openai.py](https://codecov.io/gh/huggingface/transformers/pull/5812/diff?src=pr&el=tree#diff-c3JjL3RyYW5zZm9ybWVycy9tb2RlbGluZ19vcGVuYWkucHk=) | `82.31% <0.00%> (+1.28%)` | :arrow_up: |\n| [src/transformers/modeling\\_tf\\_openai.py](https://codecov.io/gh/huggingface/transformers/pull/5812/diff?src=pr&el=tree#diff-c3JjL3RyYW5zZm9ybWVycy9tb2RlbGluZ190Zl9vcGVuYWkucHk=) | `95.18% <0.00%> (+74.91%)` | :arrow_up: |\n\n------\n\n[Continue to review full report at Codecov](https://codecov.io/gh/huggingface/transformers/pull/5812?src=pr&el=continue).\n> **Legend** - [Click here to learn more](https://docs.codecov.io/docs/codecov-delta)\n> `Δ = absolute <relative> (impact)`, `ø = not affected`, `? = missing data`\n> Powered by [Codecov](https://codecov.io/gh/huggingface/transformers/pull/5812?src=pr&el=footer). Last update [89a78be...a5f39ab](https://codecov.io/gh/huggingface/transformers/pull/5812?src=pr&el=lastupdated). Read the [comment docs](https://docs.codecov.io/docs/pull-request-comments).\n" ]

[]

[]

[ "# [Codecov](https://codecov.io/gh/huggingface/transformers/pull/5811?src=pr&el=h1) Report\n> Merging [#5811](https://codecov.io/gh/huggingface/transformers/pull/5811?src=pr&el=desc) into [master](https://codecov.io/gh/huggingface/transformers/commit/89a78be51f1c2afd263da66ec76d3297432f0c2a&el=desc) will **decrease** coverage by `0.80%`.\n> The diff coverage is `100.00%`.\n\n[](https://codecov.io/gh/huggingface/transformers/pull/5811?src=pr&el=tree)\n\n```diff\n@@ Coverage Diff @@\n## master #5811 +/- ##\n==========================================\n- Coverage 78.19% 77.38% -0.81% \n==========================================\n Files 146 146 \n Lines 26047 26047 \n==========================================\n- Hits 20367 20157 -210 \n- Misses 5680 5890 +210 \n```\n\n\n| [Impacted Files](https://codecov.io/gh/huggingface/transformers/pull/5811?src=pr&el=tree) | Coverage Δ | |\n|---|---|---|\n| [src/transformers/modeling\\_longformer.py](https://codecov.io/gh/huggingface/transformers/pull/5811/diff?src=pr&el=tree#diff-c3JjL3RyYW5zZm9ybWVycy9tb2RlbGluZ19sb25nZm9ybWVyLnB5) | `89.21% <100.00%> (ø)` | |\n| [src/transformers/modeling\\_tf\\_mobilebert.py](https://codecov.io/gh/huggingface/transformers/pull/5811/diff?src=pr&el=tree#diff-c3JjL3RyYW5zZm9ybWVycy9tb2RlbGluZ190Zl9tb2JpbGViZXJ0LnB5) | `23.38% <0.00%> (-73.39%)` | :arrow_down: |\n| [src/transformers/generation\\_tf\\_utils.py](https://codecov.io/gh/huggingface/transformers/pull/5811/diff?src=pr&el=tree#diff-c3JjL3RyYW5zZm9ybWVycy9nZW5lcmF0aW9uX3RmX3V0aWxzLnB5) | `86.21% <0.00%> (ø)` | |\n| [src/transformers/generation\\_utils.py](https://codecov.io/gh/huggingface/transformers/pull/5811/diff?src=pr&el=tree#diff-c3JjL3RyYW5zZm9ybWVycy9nZW5lcmF0aW9uX3V0aWxzLnB5) | `97.11% <0.00%> (+0.28%)` | :arrow_up: |\n| [src/transformers/modeling\\_openai.py](https://codecov.io/gh/huggingface/transformers/pull/5811/diff?src=pr&el=tree#diff-c3JjL3RyYW5zZm9ybWVycy9tb2RlbGluZ19vcGVuYWkucHk=) | `82.31% <0.00%> (+1.28%)` | :arrow_up: |\n| [src/transformers/modeling\\_tf\\_openai.py](https://codecov.io/gh/huggingface/transformers/pull/5811/diff?src=pr&el=tree#diff-c3JjL3RyYW5zZm9ybWVycy9tb2RlbGluZ190Zl9vcGVuYWkucHk=) | `95.18% <0.00%> (+74.91%)` | :arrow_up: |\n\n------\n\n[Continue to review full report at Codecov](https://codecov.io/gh/huggingface/transformers/pull/5811?src=pr&el=continue).\n> **Legend** - [Click here to learn more](https://docs.codecov.io/docs/codecov-delta)\n> `Δ = absolute <relative> (impact)`, `ø = not affected`, `? = missing data`\n> Powered by [Codecov](https://codecov.io/gh/huggingface/transformers/pull/5811?src=pr&el=footer). Last update [89a78be...25c6419](https://codecov.io/gh/huggingface/transformers/pull/5811?src=pr&el=lastupdated). Read the [comment docs](https://docs.codecov.io/docs/pull-request-comments).\n", "Great find. +1 to tests!" ]

[]

[]

[ "# [Codecov](https://codecov.io/gh/huggingface/transformers/pull/5810?src=pr&el=h1) Report\n> Merging [#5810](https://codecov.io/gh/huggingface/transformers/pull/5810?src=pr&el=desc) into [master](https://codecov.io/gh/huggingface/transformers/commit/89a78be51f1c2afd263da66ec76d3297432f0c2a&el=desc) will **decrease** coverage by `0.88%`.\n> The diff coverage is `n/a`.\n\n[](https://codecov.io/gh/huggingface/transformers/pull/5810?src=pr&el=tree)\n\n```diff\n@@ Coverage Diff @@\n## master #5810 +/- ##\n==========================================\n- Coverage 78.19% 77.30% -0.89% \n==========================================\n Files 146 146 \n Lines 26047 26047 \n==========================================\n- Hits 20367 20136 -231 \n- Misses 5680 5911 +231 \n```\n\n\n| [Impacted Files](https://codecov.io/gh/huggingface/transformers/pull/5810?src=pr&el=tree) | Coverage Δ | |\n|---|---|---|\n| [src/transformers/tokenization\\_bert.py](https://codecov.io/gh/huggingface/transformers/pull/5810/diff?src=pr&el=tree#diff-c3JjL3RyYW5zZm9ybWVycy90b2tlbml6YXRpb25fYmVydC5weQ==) | `91.32% <ø> (ø)` | |\n| [src/transformers/modeling\\_tf\\_mobilebert.py](https://codecov.io/gh/huggingface/transformers/pull/5810/diff?src=pr&el=tree#diff-c3JjL3RyYW5zZm9ybWVycy9tb2RlbGluZ190Zl9tb2JpbGViZXJ0LnB5) | `23.38% <0.00%> (-73.39%)` | :arrow_down: |\n| [src/transformers/modeling\\_tf\\_distilbert.py](https://codecov.io/gh/huggingface/transformers/pull/5810/diff?src=pr&el=tree#diff-c3JjL3RyYW5zZm9ybWVycy9tb2RlbGluZ190Zl9kaXN0aWxiZXJ0LnB5) | `64.90% <0.00%> (-33.90%)` | :arrow_down: |\n| [src/transformers/generation\\_tf\\_utils.py](https://codecov.io/gh/huggingface/transformers/pull/5810/diff?src=pr&el=tree#diff-c3JjL3RyYW5zZm9ybWVycy9nZW5lcmF0aW9uX3RmX3V0aWxzLnB5) | `86.46% <0.00%> (+0.25%)` | :arrow_up: |\n| [src/transformers/file\\_utils.py](https://codecov.io/gh/huggingface/transformers/pull/5810/diff?src=pr&el=tree#diff-c3JjL3RyYW5zZm9ybWVycy9maWxlX3V0aWxzLnB5) | `81.49% <0.00%> (+0.29%)` | :arrow_up: |\n| [src/transformers/modeling\\_openai.py](https://codecov.io/gh/huggingface/transformers/pull/5810/diff?src=pr&el=tree#diff-c3JjL3RyYW5zZm9ybWVycy9tb2RlbGluZ19vcGVuYWkucHk=) | `82.31% <0.00%> (+1.28%)` | :arrow_up: |\n| [src/transformers/modeling\\_tf\\_roberta.py](https://codecov.io/gh/huggingface/transformers/pull/5810/diff?src=pr&el=tree#diff-c3JjL3RyYW5zZm9ybWVycy9tb2RlbGluZ190Zl9yb2JlcnRhLnB5) | `93.36% <0.00%> (+49.37%)` | :arrow_up: |\n| [src/transformers/modeling\\_tf\\_openai.py](https://codecov.io/gh/huggingface/transformers/pull/5810/diff?src=pr&el=tree#diff-c3JjL3RyYW5zZm9ybWVycy9tb2RlbGluZ190Zl9vcGVuYWkucHk=) | `95.18% <0.00%> (+74.91%)` | :arrow_up: |\n\n------\n\n[Continue to review full report at Codecov](https://codecov.io/gh/huggingface/transformers/pull/5810?src=pr&el=continue).\n> **Legend** - [Click here to learn more](https://docs.codecov.io/docs/codecov-delta)\n> `Δ = absolute <relative> (impact)`, `ø = not affected`, `? = missing data`\n> Powered by [Codecov](https://codecov.io/gh/huggingface/transformers/pull/5810?src=pr&el=footer). Last update [89a78be...d7b3627](https://codecov.io/gh/huggingface/transformers/pull/5810?src=pr&el=lastupdated). Read the [comment docs](https://docs.codecov.io/docs/pull-request-comments).\n", "@LysandreJik \r\n\r\nCan you merge this PR? Thank you:)" ]

[]

[ { "login": "patrickvonplaten", "id": 23423619, "node_id": "MDQ6VXNlcjIzNDIzNjE5", "avatar_url": "https://avatars.githubusercontent.com/u/23423619?v=4", "gravatar_id": "", "url": "https://api.github.com/users/patrickvonplaten", "html_url": "https://github.com/patrickvonplaten", "followers_url": "https://api.github.com/users/patrickvonplaten/followers", "following_url": "https://api.github.com/users/patrickvonplaten/following{/other_user}", "gists_url": "https://api.github.com/users/patrickvonplaten/gists{/gist_id}", "starred_url": "https://api.github.com/users/patrickvonplaten/starred{/owner}{/repo}", "subscriptions_url": "https://api.github.com/users/patrickvonplaten/subscriptions", "organizations_url": "https://api.github.com/users/patrickvonplaten/orgs", "repos_url": "https://api.github.com/users/patrickvonplaten/repos", "events_url": "https://api.github.com/users/patrickvonplaten/events{/privacy}", "received_events_url": "https://api.github.com/users/patrickvonplaten/received_events", "type": "User", "site_admin": false } ]

[ "The `EncoderDecoderModel` does not work with longformer yet. Closing this in favor of https://github.com/huggingface/transformers/issues/4225 ." ]

[]

[]

[]

[ { "id": 1314768611, "node_id": "MDU6TGFiZWwxMzE0NzY4NjEx", "url": "https://api.github.com/repos/huggingface/transformers/labels/wontfix", "name": "wontfix", "color": "ffffff", "default": true, "description": null } ]

[]

[ "Hi @laibamehnaz , what is your batch size ?\r\n\r\nHere's my wild guess. I think one of your batch has empty lines. The batches are trimmed using the below function to remove excessive pad tokens\r\n\r\n```python3\r\ndef trim_batch(\r\n input_ids, pad_token_id, attention_mask=None,\r\n):\r\n \"\"\"Remove columns that are populated exclusively by pad_token_id\"\"\"\r\n keep_column_mask = input_ids.ne(pad_token_id).any(dim=0)\r\n if attention_mask is None:\r\n return input_ids[:, keep_column_mask]\r\n else:\r\n return (input_ids[:, keep_column_mask], attention_mask[:, keep_column_mask])\r\n```\r\n\r\nif your batch has empty lines then after tokenizing it'll contain only pad tokens, and when its passed to trim batch it'll remove all those pad tokens and the batch will be empty i.e of shape [batch_size, 0]. This could the reason for the `RuntimeError`\r\n\r\nyou can verify this with \r\n```python3\r\ntok = T5Tokenizer.from_pretrained(\"t5-base\")\r\ntext_batch = [\"\", \"\"]\r\nenc = tok(text_batch, max_length=64, padding=\"max_length\", return_tensors=\"pt\")\r\nprint(enc['input_ids'].shape)\r\nids = trim_batch(enc['input_ids'], tok.pad_token_id)\r\ninput_shape = ids.size()\r\nids = ids.view(-1, input_shape[-1])\r\n```\r\n\r\nit'll give the same `RuntimeError`.\r\n\r\nSo I think you should check your dataset for empty lines, that should solve this issue ", "Hi @patil-suraj ,\r\nThank you so much for your quick response. I have checked my dataset already for the same issue, and it didn't show any empty lines, and that's why I was so confused. But sure, I will check again. Also, my batch_size=1. ", "If your bs is 1 then most probably blank line is the reason ", "@laibamehnaz can you just iterate through the dataloader with bs 1 and see when the shape comes out to be [bs, 0] this should allow you to pinpoint the error", "Hi @patil-suraj, I tried what you suggested, but I do not get the same error. My smallest sentence is 1 word long, and there are no empty lines. ", "Interesting, can you post the word here, maybe tokenizer is skipping it then. One more thing you can try is do forward pass only with that one word example and see if it gives the error", "Also pinging @sshleifer ", "Also, there is [this line](https://github.com/huggingface/transformers/blob/ab0b3fa15c159034a55ee95ac39b7a6ba4527c4a/examples/seq2seq/finetune.py#L133) in finetune.py, which suggests that if your decoder_input_ids are length 0 and you're not adding special tokens (as was the case previously for t5), the labels would get truncated to shape (1, 0).", "https://github.com/huggingface/transformers/blob/ab0b3fa15c159034a55ee95ac39b7a6ba4527c4a/examples/seq2seq/finetune.py#L133", "Right, so one word example is encoded as single id and due to shifting and no special token, shape becomes (1,0). So one more reason to automatically add `</s>` with t5 tokenizer ", "@laibamehnaz so immediate solution for this issue is to add `</s>` at the end of every target example text", "Alright, thank you so much. Will get back after trying it out. ", "Works fine now. Thank you @patil-suraj.", "hi @laibamehnaz , there's another bug in finetune.py when using T5. See issue #5987 and PR #5994", "Fix is easy, just check the change in the PR", "Thanks a lot!", "That PR is merged now, you can now just clone the master or pull if you've already cloned", "I did notice that. Thanks a lot!", "This issue has been automatically marked as stale because it has not had recent activity. It will be closed if no further activity occurs. Thank you for your contributions.\n" ]

[ { "id": 1314768611, "node_id": "MDU6TGFiZWwxMzE0NzY4NjEx", "url": "https://api.github.com/repos/huggingface/transformers/labels/wontfix", "name": "wontfix", "color": "ffffff", "default": true, "description": null } ]

[]

[ "I don't see anything obviously wrong on our side, so you should ask the model authors (@deeppavlov and other members of @deepmipt) – maybe they uploaded the model without a trained final LM prediction layer, for instance.\r\n\r\nOff topic, but e.g. by contrast this translation model works very well for russian: https://huggingface.co/Helsinki-NLP/opus-mt-ru-en?text=%D0%A3+%D0%BC%D0%B5%D0%BD%D1%8F+%D0%B1%D1%8B%D0%BB+%D1%81%D0%B0%D0%BB%D0%B0%D1%82+%D0%BD%D0%B0+%D0%BE%D0%B1%D0%B5%D0%B4", "Hi!\r\nWe (DeepPavlov) might have a problem with Russian model that was uploaded to HugginFace and we need to check it.\r\nEverything should be okay with MLM head if you use checkpoints from here:\r\nhttp://docs.deeppavlov.ai/en/master/features/pretrained_vectors.html#downloads\r\n", "This issue has been automatically marked as stale because it has not had recent activity. It will be closed if no further activity occurs. Thank you for your contributions.\n" ]

[ { "id": 1314768611, "node_id": "MDU6TGFiZWwxMzE0NzY4NjEx", "url": "https://api.github.com/repos/huggingface/transformers/labels/wontfix", "name": "wontfix", "color": "ffffff", "default": true, "description": null } ]

[]

[ "This issue has been automatically marked as stale because it has not had recent activity. It will be closed if no further activity occurs. Thank you for your contributions.\n" ]

[]

[]

[ "@StillKeepTry Don't forget to ping me if you need another round of review!\r\n\r\nBy the way, I haven't seen any refactorization for code like this:\r\n```python\r\nfrom .modeling_bert import (\r\n BertEmbeddings, \r\n BertPreTrainedModel, \r\n BertIntermediate, \r\n BertOutput, \r\n BertSelfAttention,\r\n BertSelfOutput, \r\n BertAttention, \r\n BertModel\r\n)\r\n```\r\nI recommend you copy the code like `BertEmbeddings` and paste them into your own modeling file and rename it to `MPNetEmebedding`.", "Would you please add tests for MPNet? You can do it following other models.", "Also there are several CI fails, please fix it and we can move to the next step. Thanks for your great contribution!", "Hey @StillKeepTry - there seems to have been a problem with a git rebase. We sadly cannot review 905 changed files. Can you open a new PR that only has the necessary changes? ", "> Hey @StillKeepTry - there seems to have been a problem with a git rebase. We sadly cannot review 905 changed files. Can you open a new PR that only has the necessary changes?\r\n\r\nok", "The new PR is submitted at [https://github.com/huggingface/transformers/pull/8804](https://github.com/huggingface/transformers/pull/8804)\r\n\r\n@patrickvonplaten " ]

[]

[]

[ "I can load the pipeline locally so I am not sure what's happening here, indeed. Will post here when we know more.", "Hey @julien-c, thanks for the quick response!\r\n\r\nI'm assuming the inference API is using the `transformers-cli serve` ?\r\nWhen I do the same locally, I'm able to get results at **localhost:8888/forward** using `POST -d {\"inputs\": [\"< Text >\"]}` \r\nBut when I try it on the https://api-inference.huggingface.co/models/sampathkethineedi/industry-classification-api, it doesn't work.\r\n\r\nThe Hosted Inference API has the same structure as `transformers-cli serve` or is it any different?", "[https://huggingface.co/sampathkethineedi/industry-classification-api](https://huggingface.co/sampathkethineedi/industry-classification-api)\n\nI just checked and the API is working. Would love to know what was causing the problem. Closing the issue.", "@sampathkethineedi can i get dataset u trained on " ]

[ { "id": 1314768611, "node_id": "MDU6TGFiZWwxMzE0NzY4NjEx", "url": "https://api.github.com/repos/huggingface/transformers/labels/wontfix", "name": "wontfix", "color": "ffffff", "default": true, "description": null } ]

[]

[ "This issue has been automatically marked as stale because it has not had recent activity. It will be closed if no further activity occurs. Thank you for your contributions.\n" ]

[]

[]

[ "# [Codecov](https://codecov.io/gh/huggingface/transformers/pull/5801?src=pr&el=h1) Report\n> Merging [#5801](https://codecov.io/gh/huggingface/transformers/pull/5801?src=pr&el=desc) into [master](https://codecov.io/gh/huggingface/transformers/commit/8ce610bc96ace8be6b514c0f2c3fb0f76eb7ee15&el=desc) will **increase** coverage by `0.08%`.\n> The diff coverage is `75.00%`.\n\n[](https://codecov.io/gh/huggingface/transformers/pull/5801?src=pr&el=tree)\n\n```diff\n@@ Coverage Diff @@\n## master #5801 +/- ##\n==========================================\n+ Coverage 77.33% 77.42% +0.08% \n==========================================\n Files 146 146 \n Lines 26047 26047 \n==========================================\n+ Hits 20143 20166 +23 \n+ Misses 5904 5881 -23 \n```\n\n\n| [Impacted Files](https://codecov.io/gh/huggingface/transformers/pull/5801?src=pr&el=tree) | Coverage Δ | |\n|---|---|---|\n| [src/transformers/benchmark/benchmark\\_tf.py](https://codecov.io/gh/huggingface/transformers/pull/5801/diff?src=pr&el=tree#diff-c3JjL3RyYW5zZm9ybWVycy9iZW5jaG1hcmsvYmVuY2htYXJrX3RmLnB5) | `65.03% <50.00%> (+3.49%)` | :arrow_up: |\n| [src/transformers/benchmark/benchmark.py](https://codecov.io/gh/huggingface/transformers/pull/5801/diff?src=pr&el=tree#diff-c3JjL3RyYW5zZm9ybWVycy9iZW5jaG1hcmsvYmVuY2htYXJrLnB5) | `81.10% <100.00%> (+7.08%)` | :arrow_up: |\n| [src/transformers/modeling\\_tf\\_openai.py](https://codecov.io/gh/huggingface/transformers/pull/5801/diff?src=pr&el=tree#diff-c3JjL3RyYW5zZm9ybWVycy9tb2RlbGluZ190Zl9vcGVuYWkucHk=) | `20.27% <0.00%> (-74.92%)` | :arrow_down: |\n| [src/transformers/tokenization\\_xlm.py](https://codecov.io/gh/huggingface/transformers/pull/5801/diff?src=pr&el=tree#diff-c3JjL3RyYW5zZm9ybWVycy90b2tlbml6YXRpb25feGxtLnB5) | `16.26% <0.00%> (-66.67%)` | :arrow_down: |\n| [src/transformers/data/processors/squad.py](https://codecov.io/gh/huggingface/transformers/pull/5801/diff?src=pr&el=tree#diff-c3JjL3RyYW5zZm9ybWVycy9kYXRhL3Byb2Nlc3NvcnMvc3F1YWQucHk=) | `28.34% <0.00%> (-6.55%)` | :arrow_down: |\n| [src/transformers/generation\\_tf\\_utils.py](https://codecov.io/gh/huggingface/transformers/pull/5801/diff?src=pr&el=tree#diff-c3JjL3RyYW5zZm9ybWVycy9nZW5lcmF0aW9uX3RmX3V0aWxzLnB5) | `83.70% <0.00%> (-2.76%)` | :arrow_down: |\n| [src/transformers/modeling\\_openai.py](https://codecov.io/gh/huggingface/transformers/pull/5801/diff?src=pr&el=tree#diff-c3JjL3RyYW5zZm9ybWVycy9tb2RlbGluZ19vcGVuYWkucHk=) | `81.02% <0.00%> (-1.29%)` | :arrow_down: |\n| [src/transformers/tokenization\\_utils.py](https://codecov.io/gh/huggingface/transformers/pull/5801/diff?src=pr&el=tree#diff-c3JjL3RyYW5zZm9ybWVycy90b2tlbml6YXRpb25fdXRpbHMucHk=) | `89.95% <0.00%> (-0.81%)` | :arrow_down: |\n| ... and [1 more](https://codecov.io/gh/huggingface/transformers/pull/5801/diff?src=pr&el=tree-more) | |\n\n------\n\n[Continue to review full report at Codecov](https://codecov.io/gh/huggingface/transformers/pull/5801?src=pr&el=continue).\n> **Legend** - [Click here to learn more](https://docs.codecov.io/docs/codecov-delta)\n> `Δ = absolute <relative> (impact)`, `ø = not affected`, `? = missing data`\n> Powered by [Codecov](https://codecov.io/gh/huggingface/transformers/pull/5801?src=pr&el=footer). Last update [8ce610b...dbf09d1](https://codecov.io/gh/huggingface/transformers/pull/5801?src=pr&el=lastupdated). Read the [comment docs](https://docs.codecov.io/docs/pull-request-comments).\n", "Pinging @LysandreJik for notification.", "Nice!" ]

[]

[]

[ "Hey @almaLindborg,\r\n\r\nThis is a known warning. We should probably disable this...a couple of other models have it as well. But the model works as intended, so no worries about the message.\r\n\r\nAlso pinging @sshleifer for notification." ]

[ { "id": 1314768611, "node_id": "MDU6TGFiZWwxMzE0NzY4NjEx", "url": "https://api.github.com/repos/huggingface/transformers/labels/wontfix", "name": "wontfix", "color": "ffffff", "default": true, "description": null } ]

[]

[ "Cannot reproduce. The command:\r\n\r\n```python \r\nfrom transformers import AutoModelWithLMHead\r\nmodel = AutoModelWithLMHead.from_pretrained(\"bert-base-chinese\")\r\n```\r\n\r\nworks fine on master. Can you update to v3.0.2 `pip install --upgrade transformers` and check again? :-) ", "> Cannot reproduce. The command:\r\n> \r\n> ```python\r\n> from transformers import AutoModelWithLMHead\r\n> model = AutoModelWithLMHead.from_pretrained(\"bert-base-chinese\")\r\n> ```\r\n> \r\n> works fine on master. Can you update to v3.0.2 `pip install --upgrade transformers` and check again? :-)\r\n\r\nStill not work for me. I tried to download the weights directly, but another error occurred...\r\nUnicodeDecodeError: 'utf-8' codec can't decode byte 0x80 in position 0: invalid start byte", "Can you please post your environment info here `python src/transformers/commands/env.py`?", "I'm seeing the same error when trying to load a GTP2 checkpoint model (using both `GPT2Model` and `AutoModel`):\r\n\r\n```\r\nmodel = GPT2Model.from_pretrained('./test_01/test_01.index', from_tf=True) # throws UnicodeDecodeError\r\nmodel = GPT2Model.from_pretrained('./test_01/test_01.index') # throws UnicodeDecodeError\r\nmodel = AutoModel.from_pretrained('./test_01/test_01.index', from_tf=True) # throws UnicodeDecodeError\r\nmodel = AutoModel.from_pretrained('./test_01/test_01.index') # throws UnicodeDecodeError\r\n```\r\n\r\nI could probably try every possible variation of loading that model and hit the same error.\r\n\r\nI've also used checkpoint models that, in theory, should work.\r\n\r\nIf I use `GPT2LMHeadModel.from_pretrained('gpt2-medium')` (or any thing that allows me to load a model by name) it works fine.\r\n\r\nMy env:\r\n\r\n```\r\n- `transformers` version: 3.0.2\r\n- Platform: macOS-10.15.5-x86_64-i386-64bit\r\n- Python version: 3.8.3\r\n- PyTorch version (GPU?): 1.6.0 (False)\r\n- Tensorflow version (GPU?): 2.3.0 (False)\r\n- Using GPU in script?: No\r\n- Using distributed or parallel set-up in script?: Not sure\r\n```\r\n", "@elbowdonkey, \r\n\r\ncan you try just running:\r\n```python \r\nmodel = GPT2Model.from_pretrained('./test_01/\", from_tf=True)\r\n```\r\n\r\nwhere the relevant files can be found in `test_01`?", "I get a different error:\r\n\r\n```python\r\nmodel = GPT2Model.from_pretrained(\"./test_02/\", from_tf=True)\r\n2020-08-10 12:32:10.047629: I tensorflow/core/platform/cpu_feature_guard.cc:142] This TensorFlow binary is optimized with oneAPI Deep Neural Network Library (oneDNN)to use the following CPU instructions in performance-critical operations: AVX2 FMA\r\nTo enable them in other operations, rebuild TensorFlow with the appropriate compiler flags.\r\n2020-08-10 12:32:10.064526: I tensorflow/compiler/xla/service/service.cc:168] XLA service 0x7f8eb76b8f80 initialized for platform Host (this does not guarantee that XLA will be used). Devices:\r\n2020-08-10 12:32:10.064543: I tensorflow/compiler/xla/service/service.cc:176] StreamExecutor device (0): Host, Default Version\r\n2020-08-10 12:32:10.071086: W tensorflow/python/util/util.cc:348] Sets are not currently considered sequences, but this may change in the future, so consider avoiding using them.\r\n2020-08-10 12:32:27.584426: W tensorflow/core/util/tensor_slice_reader.cc:95] Could not open ./test_02/pytorch_model.bin: Data loss: not an sstable (bad magic number): perhaps your file is in a different file format and you need to use a different restore operator?\r\nTraceback (most recent call last):\r\n File \"<stdin>\", line 1, in <module>\r\n File \"/Users/--/.pyenv/versions/3.8.3/lib/python3.8/site-packages/transformers/modeling_utils.py\", line 808, in from_pretrained\r\n model = load_tf2_checkpoint_in_pytorch_model(model, resolved_archive_file, allow_missing_keys=True)\r\n File \"/Users/--/.pyenv/versions/3.8.3/lib/python3.8/site-packages/transformers/modeling_tf_pytorch_utils.py\", line 261, in load_tf2_checkpoint_in_pytorch_model\r\n tf_model.load_weights(tf_checkpoint_path, by_name=True)\r\n File \"/Users/--/.pyenv/versions/3.8.3/lib/python3.8/site-packages/tensorflow/python/keras/engine/training.py\", line 2204, in load_weights\r\n with h5py.File(filepath, 'r') as f:\r\n File \"/Users/--/.pyenv/versions/3.8.3/lib/python3.8/site-packages/h5py/_hl/files.py\", line 406, in __init__\r\n fid = make_fid(name, mode, userblock_size,\r\n File \"/Users/--/.pyenv/versions/3.8.3/lib/python3.8/site-packages/h5py/_hl/files.py\", line 173, in make_fid\r\n fid = h5f.open(name, flags, fapl=fapl)\r\n File \"h5py/_objects.pyx\", line 54, in h5py._objects.with_phil.wrapper\r\n File \"h5py/_objects.pyx\", line 55, in h5py._objects.with_phil.wrapper\r\n File \"h5py/h5f.pyx\", line 88, in h5py.h5f.open\r\nOSError: Unable to open file (file signature not found)\r\n```\r\n\r\nThe model I'm trying to use is a model that was converted from a checkpoint to a pytorch model. I have no idea what kind of checkpoint model it was (it has several files: `checkpoint` and `vocab.bpe`, `hparams.json`, and `test_02.data-00000-of-00001`, among others.)", "This issue has been automatically marked as stale because it has not had recent activity. It will be closed if no further activity occurs. Thank you for your contributions.\n", "Hi,\r\n\r\nI would like to request that this ticket be opened back up. I'm having the same issue but with the default pretrained model:\r\n\r\n```\r\n---------------------------------------------------------------------------\r\nOSError Traceback (most recent call last)\r\n~\\Anaconda3\\lib\\site-packages\\transformers\\modeling_utils.py in from_pretrained(cls, pretrained_model_name_or_path, *model_args, **kwargs)\r\n 654 if resolved_archive_file is None:\r\n--> 655 raise EnvironmentError\r\n 656 except EnvironmentError:\r\n\r\nOSError: \r\n\r\nDuring handling of the above exception, another exception occurred:\r\n\r\nOSError Traceback (most recent call last)\r\n<ipython-input-81-7e9fe0224671> in <module>\r\n 3 #Load AutoModel from huggingface model repository\r\n 4 tokenizer = AutoTokenizer.from_pretrained(\"sentence-transformers/bert-base-nli-mean-tokens\")\r\n----> 5 model = AutoModel.from_pretrained(\"sentence-transformers/bert-base-nli-mean-tokens\", from_tf=True)\r\n\r\n~\\Anaconda3\\lib\\site-packages\\transformers\\modeling_auto.py in from_pretrained(cls, pretrained_model_name_or_path, *model_args, **kwargs)\r\n 500 for config_class, model_class in MODEL_MAPPING.items():\r\n 501 if isinstance(config, config_class):\r\n--> 502 return model_class.from_pretrained(pretrained_model_name_or_path, *model_args, config=config, **kwargs)\r\n 503 raise ValueError(\r\n 504 \"Unrecognized configuration class {} for this kind of AutoModel: {}.\\n\"\r\n\r\n~\\Anaconda3\\lib\\site-packages\\transformers\\modeling_utils.py in from_pretrained(cls, pretrained_model_name_or_path, *model_args, **kwargs)\r\n 660 f\"- or '{pretrained_model_name_or_path}' is the correct path to a directory containing a file named one of {WEIGHTS_NAME}, {TF2_WEIGHTS_NAME}, {TF_WEIGHTS_NAME}.\\n\\n\"\r\n 661 )\r\n--> 662 raise EnvironmentError(msg)\r\n 663 \r\n 664 if resolved_archive_file == archive_file:\r\n\r\nOSError: Can't load weights for 'sentence-transformers/bert-base-nli-mean-tokens'. Make sure that:\r\n\r\n- 'sentence-transformers/bert-base-nli-mean-tokens' is a correct model identifier listed on 'https://huggingface.co/models'\r\n\r\n- or 'sentence-transformers/bert-base-nli-mean-tokens' is the correct path to a directory containing a file named one of pytorch_model.bin, tf_model.h5, model.ckpt.\r\n```\r\n\r\n**To reproduce:**\r\n```\r\nfrom transformers import AutoTokenizer, AutoModel\r\n\r\n#Load AutoModel from huggingface model repository\r\ntokenizer = AutoTokenizer.from_pretrained(\"sentence-transformers/bert-base-nli-mean-tokens\")\r\nmodel = AutoModel.from_pretrained(\"sentence-transformers/bert-base-nli-mean-tokens\", from_tf=True)\r\n```\r\n\r\nMy env:\r\n\r\n- `transformers` version: 3.0.2\r\n- Platform: Window 10 Enterprise, version 1909, 16GB RAM, 64 Bit OS, x64-based processor\r\n- Python version: 3.8.3\r\n- Torch version: 1.6.0+cpu", "Hey @Ecanlilar,\r\n\r\nThis model exists only in PT so either you do:\r\n\r\n```python\r\nfrom transformers import AutoTokenizer, AutoModel\r\n\r\n#Load AutoModel from huggingface model repository\r\ntokenizer = AutoTokenizer.from_pretrained(\"sentence-transformers/bert-base-nli-mean-tokens\")\r\nmodel = AutoModel.from_pretrained(\"sentence-transformers/bert-base-nli-mean-tokens\")\r\n```\r\n\r\nor \r\n\r\n```python\r\nfrom transformers import AutoTokenizer, TFAutoModel\r\n\r\n#Load AutoModel from huggingface model repository\r\ntokenizer = AutoTokenizer.from_pretrained(\"sentence-transformers/bert-base-nli-mean-tokens\")\r\nmodel = TFAutoModel.from_pretrained(\"sentence-transformers/bert-base-nli-mean-tokens\", from_pt=True)\r\n```", "> Hey @Ecanlilar,\r\n> \r\n> This model exists only in PT so either you do:\r\n> \r\n> ```python\r\n> from transformers import AutoTokenizer, AutoModel\r\n> \r\n> #Load AutoModel from huggingface model repository\r\n> tokenizer = AutoTokenizer.from_pretrained(\"sentence-transformers/bert-base-nli-mean-tokens\")\r\n> model = AutoModel.from_pretrained(\"sentence-transformers/bert-base-nli-mean-tokens\")\r\n> ```\r\n> \r\n> or\r\n> \r\n> ```python\r\n> from transformers import AutoTokenizer, TFAutoModel\r\n> \r\n> #Load AutoModel from huggingface model repository\r\n> tokenizer = AutoTokenizer.from_pretrained(\"sentence-transformers/bert-base-nli-mean-tokens\")\r\n> model = TFAutoModel.from_pretrained(\"sentence-transformers/bert-base-nli-mean-tokens\", from_pt=True)\r\n> ```\r\n\r\nThis isn't working for me. I'm using the latest version of the transformers library (4.10.2). I'm getting the same error as Ecanlilar." ]

[]

[ { "login": "sshleifer", "id": 6045025, "node_id": "MDQ6VXNlcjYwNDUwMjU=", "avatar_url": "https://avatars.githubusercontent.com/u/6045025?v=4", "gravatar_id": "", "url": "https://api.github.com/users/sshleifer", "html_url": "https://github.com/sshleifer", "followers_url": "https://api.github.com/users/sshleifer/followers", "following_url": "https://api.github.com/users/sshleifer/following{/other_user}", "gists_url": "https://api.github.com/users/sshleifer/gists{/gist_id}", "starred_url": "https://api.github.com/users/sshleifer/starred{/owner}{/repo}", "subscriptions_url": "https://api.github.com/users/sshleifer/subscriptions", "organizations_url": "https://api.github.com/users/sshleifer/orgs", "repos_url": "https://api.github.com/users/sshleifer/repos", "events_url": "https://api.github.com/users/sshleifer/events{/privacy}", "received_events_url": "https://api.github.com/users/sshleifer/received_events", "type": "User", "site_admin": false } ]

[ "@sshleifer have any guidance on these two errors?\r\n\r\n\r\n## T5\r\n\r\n```python\r\n________________ test_finetune[patrickvonplaten/t5-tiny-random] ________________\r\n[gw3] linux -- Python 3.6.11 /usr/local/bin/python\r\n\r\nmodel = 'patrickvonplaten/t5-tiny-random'\r\n\r\n @pytest.mark.parametrize(\r\n [\"model\"], [pytest.param(T5_TINY), pytest.param(BART_TINY), pytest.param(MBART_TINY), pytest.param(MARIAN_TINY)]\r\n )\r\n def test_finetune(model):\r\n args_d: dict = CHEAP_ARGS.copy()\r\n task = \"translation\" if model in [MBART_TINY, MARIAN_TINY] else \"summarization\"\r\n tmp_dir = make_test_data_dir()\r\n output_dir = tempfile.mkdtemp(prefix=\"output_\")\r\n args_d.update(\r\n data_dir=tmp_dir,\r\n model_name_or_path=model,\r\n tokenizer_name=None,\r\n train_batch_size=2,\r\n eval_batch_size=2,\r\n output_dir=output_dir,\r\n do_predict=True,\r\n task=task,\r\n src_lang=\"en_XX\",\r\n tgt_lang=\"ro_RO\",\r\n freeze_encoder=True,\r\n freeze_embeds=True,\r\n )\r\n assert \"n_train\" in args_d\r\n args = argparse.Namespace(**args_d)\r\n> module = main(args)\r\n\r\nexamples/seq2seq/test_seq2seq_examples.py:233: \r\n_ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ \r\nexamples/seq2seq/finetune.py:298: in main\r\n model: SummarizationModule = SummarizationModule(args)\r\nexamples/seq2seq/finetune.py:95: in __init__\r\n freeze_params(self.model.model.encoder) # TODO: this will break for t5\r\n_ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ \r\n\r\nself = T5ForConditionalGeneration(\r\n (shared): Embedding(32128, 64)\r\n (encoder): T5Stack(\r\n (embed_tokens): Embedding(32128...\r\n (dropout): Dropout(p=0.1, inplace=False)\r\n )\r\n (lm_head): Linear(in_features=64, out_features=32128, bias=False)\r\n)\r\nname = 'model'\r\n\r\n def __getattr__(self, name):\r\n if '_parameters' in self.__dict__:\r\n _parameters = self.__dict__['_parameters']\r\n if name in _parameters:\r\n return _parameters[name]\r\n if '_buffers' in self.__dict__:\r\n _buffers = self.__dict__['_buffers']\r\n if name in _buffers:\r\n return _buffers[name]\r\n if '_modules' in self.__dict__:\r\n modules = self.__dict__['_modules']\r\n if name in modules:\r\n return modules[name]\r\n raise AttributeError(\"'{}' object has no attribute '{}'\".format(\r\n> type(self).__name__, name))\r\nE AttributeError: 'T5ForConditionalGeneration' object has no attribute 'model'\r\n\r\n/usr/local/lib/python3.6/site-packages/torch/nn/modules/module.py:594: AttributeError\r\n```\r\n\r\n## MBart\r\n```python\r\n_____________________ test_finetune[sshleifer/tiny-mbart] ______________________\r\n[gw3] linux -- Python 3.6.11 /usr/local/bin/python\r\n\r\nmodel = 'sshleifer/tiny-mbart'\r\n\r\n @pytest.mark.parametrize(\r\n [\"model\"], [pytest.param(T5_TINY), pytest.param(BART_TINY), pytest.param(MBART_TINY), pytest.param(MARIAN_TINY)]\r\n )\r\n def test_finetune(model):\r\n args_d: dict = CHEAP_ARGS.copy()\r\n task = \"translation\" if model in [MBART_TINY, MARIAN_TINY] else \"summarization\"\r\n tmp_dir = make_test_data_dir()\r\n output_dir = tempfile.mkdtemp(prefix=\"output_\")\r\n args_d.update(\r\n data_dir=tmp_dir,\r\n model_name_or_path=model,\r\n tokenizer_name=None,\r\n train_batch_size=2,\r\n eval_batch_size=2,\r\n output_dir=output_dir,\r\n do_predict=True,\r\n task=task,\r\n src_lang=\"en_XX\",\r\n tgt_lang=\"ro_RO\",\r\n freeze_encoder=True,\r\n freeze_embeds=True,\r\n )\r\n assert \"n_train\" in args_d\r\n args = argparse.Namespace(**args_d)\r\n> module = main(args)\r\n\r\nexamples/seq2seq/test_seq2seq_examples.py:233: \r\n_ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ \r\nexamples/seq2seq/finetune.py:324: in main\r\n logger=logger,\r\nexamples/lightning_base.py:312: in generic_train\r\n trainer.fit(model)\r\n/usr/local/lib/python3.6/site-packages/pytorch_lightning/trainer/trainer.py:1038: in fit\r\n model.setup('fit')\r\nexamples/lightning_base.py:125: in setup\r\n dataloader = self.get_dataloader(\"train\", train_batch_size)\r\nexamples/seq2seq/finetune.py:193: in get_dataloader\r\n dataset = self.get_dataset(type_path)\r\nexamples/seq2seq/finetune.py:188: in get_dataset\r\n **self.dataset_kwargs,\r\n_ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ \r\n\r\nself = <seq2seq.utils.SummarizationDataset object at 0x7ff21a4592e8>\r\ntokenizer = <transformers.tokenization_bart.MBartTokenizer object at 0x7ff21f7c0b00>\r\ndata_dir = PosixPath('/tmp/tmpmc70afs6'), type_path = 'train'\r\nmax_source_length = 12, max_target_length = 12, n_obs = None\r\noverwrite_cache = False, prefix = '', src_lang = None, tgt_lang = None\r\n\r\n def __init__(\r\n self,\r\n tokenizer,\r\n data_dir,\r\n type_path=\"train\",\r\n max_source_length=1024,\r\n max_target_length=56,\r\n n_obs=None,\r\n overwrite_cache=False,\r\n prefix=\"\",\r\n src_lang=None,\r\n tgt_lang=None,\r\n ):\r\n super().__init__()\r\n # FIXME: the rstrip logic strips all the chars, it seems.\r\n tok_name = tokenizer.__class__.__name__.lower().rstrip(\"tokenizer\")\r\n if hasattr(tokenizer, \"set_lang\") and src_lang is not None:\r\n tokenizer.set_lang(src_lang) # HACK: only applies to mbart\r\n self.source = encode_file(\r\n tokenizer,\r\n os.path.join(data_dir, type_path + \".source\"),\r\n max_source_length,\r\n overwrite_cache=overwrite_cache,\r\n prefix=prefix,\r\n tok_name=tok_name,\r\n )\r\n tgt_path = os.path.join(data_dir, type_path + \".target\")\r\n if hasattr(tokenizer, \"set_lang\"):\r\n> assert tgt_lang is not None, \"--tgt_lang must be passed to build a translation\"\r\nE AssertionError: --tgt_lang must be passed to build a translation\r\n\r\nexamples/seq2seq/utils.py:112: AssertionError\r\n```", "# [Codecov](https://codecov.io/gh/huggingface/transformers/pull/5798?src=pr&el=h1) Report\n> Merging [#5798](https://codecov.io/gh/huggingface/transformers/pull/5798?src=pr&el=desc) into [master](https://codecov.io/gh/huggingface/transformers/commit/615be03f9d961c0c9722fe10e7830e011066772e&el=desc) will **decrease** coverage by `0.18%`.\n> The diff coverage is `n/a`.\n\n[](https://codecov.io/gh/huggingface/transformers/pull/5798?src=pr&el=tree)\n\n```diff\n@@ Coverage Diff @@\n## master #5798 +/- ##\n==========================================\n- Coverage 78.66% 78.48% -0.19% \n==========================================\n Files 146 146 \n Lines 26200 26200 \n==========================================\n- Hits 20611 20563 -48 \n- Misses 5589 5637 +48 \n```\n\n\n| [Impacted Files](https://codecov.io/gh/huggingface/transformers/pull/5798?src=pr&el=tree) | Coverage Δ | |\n|---|---|---|\n| [src/transformers/modeling\\_tf\\_electra.py](https://codecov.io/gh/huggingface/transformers/pull/5798/diff?src=pr&el=tree#diff-c3JjL3RyYW5zZm9ybWVycy9tb2RlbGluZ190Zl9lbGVjdHJhLnB5) | `26.02% <0.00%> (-69.52%)` | :arrow_down: |\n| [src/transformers/generation\\_tf\\_utils.py](https://codecov.io/gh/huggingface/transformers/pull/5798/diff?src=pr&el=tree#diff-c3JjL3RyYW5zZm9ybWVycy9nZW5lcmF0aW9uX3RmX3V0aWxzLnB5) | `86.21% <0.00%> (-0.51%)` | :arrow_down: |\n| [src/transformers/modeling\\_tf\\_distilbert.py](https://codecov.io/gh/huggingface/transformers/pull/5798/diff?src=pr&el=tree#diff-c3JjL3RyYW5zZm9ybWVycy9tb2RlbGluZ190Zl9kaXN0aWxiZXJ0LnB5) | `98.79% <0.00%> (+33.89%)` | :arrow_up: |\n\n------\n\n[Continue to review full report at Codecov](https://codecov.io/gh/huggingface/transformers/pull/5798?src=pr&el=continue).\n> **Legend** - [Click here to learn more](https://docs.codecov.io/docs/codecov-delta)\n> `Δ = absolute <relative> (impact)`, `ø = not affected`, `? = missing data`\n> Powered by [Codecov](https://codecov.io/gh/huggingface/transformers/pull/5798?src=pr&el=footer). Last update [615be03...ee864a0](https://codecov.io/gh/huggingface/transformers/pull/5798?src=pr&el=lastupdated). Read the [comment docs](https://docs.codecov.io/docs/pull-request-comments).\n", "Merging this now.\r\ncc @moscow25 this bumps us to `pytorch_lightning==0.8.5`, let us know if any issues.\r\ncc @clmnt , @patil-suraj, @williamFalcon \r\n\r\nThanks for the big PR @nateraw and @williamFalcon !", "Thanks @sshleifer -- `0.8.5` has been good for us this week. Much appreciated. " ]

[]

[]

[ "I have find anwser in [code](https://github.com/huggingface/transformers/blob/master/src/transformers/modeling_bert.py#L1138)." ]

[]

[]

[ "# [Codecov](https://codecov.io/gh/huggingface/transformers/pull/5796?src=pr&el=h1) Report\n> Merging [#5796](https://codecov.io/gh/huggingface/transformers/pull/5796?src=pr&el=desc) into [master](https://codecov.io/gh/huggingface/transformers/commit/7214954db42ec96603ea596c5f68b16f574fba89&el=desc) will **increase** coverage by `0.42%`.\n> The diff coverage is `100.00%`.\n\n[](https://codecov.io/gh/huggingface/transformers/pull/5796?src=pr&el=tree)\n\n```diff\n@@ Coverage Diff @@\n## master #5796 +/- ##\n==========================================\n+ Coverage 78.38% 78.80% +0.42% \n==========================================\n Files 146 146 \n Lines 26318 26318 \n==========================================\n+ Hits 20629 20741 +112 \n+ Misses 5689 5577 -112 \n```\n\n\n| [Impacted Files](https://codecov.io/gh/huggingface/transformers/pull/5796?src=pr&el=tree) | Coverage Δ | |\n|---|---|---|\n| [src/transformers/configuration\\_encoder\\_decoder.py](https://codecov.io/gh/huggingface/transformers/pull/5796/diff?src=pr&el=tree#diff-c3JjL3RyYW5zZm9ybWVycy9jb25maWd1cmF0aW9uX2VuY29kZXJfZGVjb2Rlci5weQ==) | `100.00% <100.00%> (ø)` | |\n| [src/transformers/modeling\\_encoder\\_decoder.py](https://codecov.io/gh/huggingface/transformers/pull/5796/diff?src=pr&el=tree#diff-c3JjL3RyYW5zZm9ybWVycy9tb2RlbGluZ19lbmNvZGVyX2RlY29kZXIucHk=) | `92.20% <100.00%> (ø)` | |\n| [src/transformers/modeling\\_marian.py](https://codecov.io/gh/huggingface/transformers/pull/5796/diff?src=pr&el=tree#diff-c3JjL3RyYW5zZm9ybWVycy9tb2RlbGluZ19tYXJpYW4ucHk=) | `90.90% <100.00%> (ø)` | |\n| [src/transformers/modeling\\_mobilebert.py](https://codecov.io/gh/huggingface/transformers/pull/5796/diff?src=pr&el=tree#diff-c3JjL3RyYW5zZm9ybWVycy9tb2RlbGluZ19tb2JpbGViZXJ0LnB5) | `89.45% <100.00%> (ø)` | |\n| [src/transformers/modeling\\_tf\\_electra.py](https://codecov.io/gh/huggingface/transformers/pull/5796/diff?src=pr&el=tree#diff-c3JjL3RyYW5zZm9ybWVycy9tb2RlbGluZ190Zl9lbGVjdHJhLnB5) | `95.53% <100.00%> (ø)` | |\n| [src/transformers/testing\\_utils.py](https://codecov.io/gh/huggingface/transformers/pull/5796/diff?src=pr&el=tree#diff-c3JjL3RyYW5zZm9ybWVycy90ZXN0aW5nX3V0aWxzLnB5) | `72.72% <100.00%> (ø)` | |\n| [src/transformers/tokenization\\_auto.py](https://codecov.io/gh/huggingface/transformers/pull/5796/diff?src=pr&el=tree#diff-c3JjL3RyYW5zZm9ybWVycy90b2tlbml6YXRpb25fYXV0by5weQ==) | `97.77% <100.00%> (ø)` | |\n| [src/transformers/modeling\\_tf\\_roberta.py](https://codecov.io/gh/huggingface/transformers/pull/5796/diff?src=pr&el=tree#diff-c3JjL3RyYW5zZm9ybWVycy9tb2RlbGluZ190Zl9yb2JlcnRhLnB5) | `43.98% <0.00%> (-49.38%)` | :arrow_down: |\n| [src/transformers/generation\\_tf\\_utils.py](https://codecov.io/gh/huggingface/transformers/pull/5796/diff?src=pr&el=tree#diff-c3JjL3RyYW5zZm9ybWVycy9nZW5lcmF0aW9uX3RmX3V0aWxzLnB5) | `86.21% <0.00%> (ø)` | |\n| [src/transformers/generation\\_utils.py](https://codecov.io/gh/huggingface/transformers/pull/5796/diff?src=pr&el=tree#diff-c3JjL3RyYW5zZm9ybWVycy9nZW5lcmF0aW9uX3V0aWxzLnB5) | `97.11% <0.00%> (+0.28%)` | :arrow_up: |\n| ... and [3 more](https://codecov.io/gh/huggingface/transformers/pull/5796/diff?src=pr&el=tree-more) | |\n\n------\n\n[Continue to review full report at Codecov](https://codecov.io/gh/huggingface/transformers/pull/5796?src=pr&el=continue).\n> **Legend** - [Click here to learn more](https://docs.codecov.io/docs/codecov-delta)\n> `Δ = absolute <relative> (impact)`, `ø = not affected`, `? = missing data`\n> Powered by [Codecov](https://codecov.io/gh/huggingface/transformers/pull/5796?src=pr&el=footer). Last update [7214954...39e85f8](https://codecov.io/gh/huggingface/transformers/pull/5796?src=pr&el=lastupdated). Read the [comment docs](https://docs.codecov.io/docs/pull-request-comments).\n" ]

[]

[]

[ "Yes! This will be added when starting the `Longformer` Encoder framework :-) Closing this in favor of https://github.com/huggingface/transformers/issues/5170 and https://github.com/huggingface/transformers/issues/4225" ]

[]

[]

[ "Hey @zanderbush, we are trying to move special feature requests / research questions to https://discuss.huggingface.co/ - would you mind posting it there again? " ]

[]

[ { "login": "LysandreJik", "id": 30755778, "node_id": "MDQ6VXNlcjMwNzU1Nzc4", "avatar_url": "https://avatars.githubusercontent.com/u/30755778?v=4", "gravatar_id": "", "url": "https://api.github.com/users/LysandreJik", "html_url": "https://github.com/LysandreJik", "followers_url": "https://api.github.com/users/LysandreJik/followers", "following_url": "https://api.github.com/users/LysandreJik/following{/other_user}", "gists_url": "https://api.github.com/users/LysandreJik/gists{/gist_id}", "starred_url": "https://api.github.com/users/LysandreJik/starred{/owner}{/repo}", "subscriptions_url": "https://api.github.com/users/LysandreJik/subscriptions", "organizations_url": "https://api.github.com/users/LysandreJik/orgs", "repos_url": "https://api.github.com/users/LysandreJik/repos", "events_url": "https://api.github.com/users/LysandreJik/events{/privacy}", "received_events_url": "https://api.github.com/users/LysandreJik/received_events", "type": "User", "site_admin": false } ]

[ "Thank you so much!!! I really appreciate you offering to help too!\r\n\r\nTo cover what you mentioned quickly, I can upload the model weights tomorrow! For updating the documentation, that should be no problem either. I think for the model outputs regarding this model, there probably is quite a bit of information to return (pooled output, hidden states, and attentions for the language, vision, and cross-modality encoders). I can see about adding these tomorrow and get your thoughts on that. And lastly, I am glad I can help with adding more model heads! \r\n\r\nI have a bit more testing to do especially with the tensorflow model, but i will see what I can get done and let you know if I run into any blockers or questions. Look forward to getting back to you!", "Hi, sorry for the slight delay. I added a new dataclass for Lxmert outputs, added the model card, finished all tests for the torch model, among a couple of other things. \r\n\r\nFor the tests, I have forgone the ModelMixin parent as I have found that some of the tests are hard to apply to lxmert. I have also temporarily forgone adding example commands for lxmert. Is this a neccesity, or would it be alright to leave these out?\r\n\r\nI think I updated the documentation to the new standard, but if there are still some errors, any help would be appreciated!\r\n\r\nI am running a donwstream task with the pretrained weights right now to ensure that the results are still the same! I will get back to you with these in about a day!", "Hi thank you so much for a really quick response and review! In my next commit, Ill implement the following changes and suggestions. If it wouldn't be to much to ask, I think I could actually do with much help adding the mixin tester. Given that there are quite a few tests that seem lxmert seems to be incompatable with (for example I think one of the tests required that the LxmertConfig had the 'num_hidden_layers' attribute, which wouldn't be applicable since I suppose instead, we let the user specify the number of hidden layers for each of the visual, language, and cross-modality encoders. You're judgement on deciding what makes sense with regards to Lxmert test compatibility is probably greater than mine. \r\n\r\nAlso just one last implementation detail that I should probably bring up is that for the output of the cross-modality encoder attentions, I only output the attentions when the language-hidden states are used as the input to the cross attention. Since this encoder is used for bidirectional attention, I do not store the attentions when the visual-hidden states are used as the input. Also For every layer of the cross attention encoder, each of the visual and language states are further processed by a separate self-attention layer, and I do not keep track of the attention outputs for these either. The only reason I keep track of the language attentions when used as input to the cross-modality attention layer is because it is those hidden states that are used for downstream pooling. \r\n\r\nI have yet to change the output for the TFLxmertModeling classes, which I will probably add in the commit following the one that addresses your review, but it probably would save me some time if you were able to cover that and would be greatly appreciated aswell!", "Commited the above changes, I need to do some testing with the conversion script again, but besides that. I will wait to change the tensorflow code and the modeling tests for your response. If I do make a change in my next commit to modeling_tf_lxmert, it will just be to the documentation", "I also just added some functionality to edit the number of question answering labels, simialr to the \"resize_output_embeddings\" utility in the LxmertPretrainedModel class. However, what I added seems a lot less complex than the process for resizing embeddings, so if you could take a look at these new functions, that would be awesome! I added some tests, and they do seem to work.", "Hi! Thanks for the thorough explanation. I'm adding the common tests to the tests, and will report here. May I push directly on your fork?", "Actually, since I'm bound to make changes to some of the files, I'd like you to review the changes I'm making so that we may discuss. I'll open a PR on your branch when it's in a good state.", "Okay sounds good! the output hidden states argument should be added in\nabout 30 minutes, and then I can see about adding the TF changes in the\nnext hour.\n\nOn Mon, Aug 10, 2020 at 8:12 AM Lysandre Debut <[email protected]>\nwrote:\n\n> Actually, since I'm bound to make changes to some of the files, I'd like\n> you to review the changes I'm making so that we may discuss. I'll open a PR\n> on your branch when it's in a good state.\n>\n> —\n> You are receiving this because you authored the thread.\n> Reply to this email directly, view it on GitHub\n> <https://github.com/huggingface/transformers/pull/5793#issuecomment-671318378>,\n> or unsubscribe\n> <https://github.com/notifications/unsubscribe-auth/ADLBORYS3OXI4RBT4LHTJLLR77P3FANCNFSM4O3KPDOA>\n> .\n>\n", "Great, I'll take a look at doing the Mixin tomorrow once your changes are up!", "the lxmert model in pytorch should be ready to go!", "Just pushed the same changes for TF as the PyTorch ones in https://github.com/eltoto1219/transformers/pull/1, alongside docs changes and a few patches to the PyTorch version. \r\n\r\nI'll take care of the merge commit once everything is done, it's due to `isort==5` and `black==20.8b` being released.", "# [Codecov](https://codecov.io/gh/huggingface/transformers/pull/5793?src=pr&el=h1) Report\n> Merging [#5793](https://codecov.io/gh/huggingface/transformers/pull/5793?src=pr&el=desc) into [master](https://codecov.io/gh/huggingface/transformers/commit/930153e7d2d658267b7630a047a4bfc85b86042d?el=desc) will **decrease** coverage by `1.74%`.\n> The diff coverage is `79.63%`.\n\n[](https://codecov.io/gh/huggingface/transformers/pull/5793?src=pr&el=tree)\n\n```diff\n@@ Coverage Diff @@\n## master #5793 +/- ##\n==========================================\n- Coverage 79.36% 77.62% -1.75% \n==========================================\n Files 157 161 +4 \n Lines 28569 29816 +1247 \n==========================================\n+ Hits 22675 23144 +469 \n- Misses 5894 6672 +778 \n```\n\n\n| [Impacted Files](https://codecov.io/gh/huggingface/transformers/pull/5793?src=pr&el=tree) | Coverage Δ | |\n|---|---|---|\n| [src/transformers/commands/convert.py](https://codecov.io/gh/huggingface/transformers/pull/5793/diff?src=pr&el=tree#diff-c3JjL3RyYW5zZm9ybWVycy9jb21tYW5kcy9jb252ZXJ0LnB5) | `27.58% <0.00%> (-1.51%)` | :arrow_down: |\n| [src/transformers/modeling\\_lxmert.py](https://codecov.io/gh/huggingface/transformers/pull/5793/diff?src=pr&el=tree#diff-c3JjL3RyYW5zZm9ybWVycy9tb2RlbGluZ19seG1lcnQucHk=) | `70.01% <70.01%> (ø)` | |\n| [src/transformers/modeling\\_tf\\_lxmert.py](https://codecov.io/gh/huggingface/transformers/pull/5793/diff?src=pr&el=tree#diff-c3JjL3RyYW5zZm9ybWVycy9tb2RlbGluZ190Zl9seG1lcnQucHk=) | `88.31% <88.31%> (ø)` | |\n| [src/transformers/\\_\\_init\\_\\_.py](https://codecov.io/gh/huggingface/transformers/pull/5793/diff?src=pr&el=tree#diff-c3JjL3RyYW5zZm9ybWVycy9fX2luaXRfXy5weQ==) | `99.30% <100.00%> (+0.01%)` | :arrow_up: |\n| [src/transformers/configuration\\_auto.py](https://codecov.io/gh/huggingface/transformers/pull/5793/diff?src=pr&el=tree#diff-c3JjL3RyYW5zZm9ybWVycy9jb25maWd1cmF0aW9uX2F1dG8ucHk=) | `93.33% <100.00%> (+0.15%)` | :arrow_up: |\n| [src/transformers/configuration\\_lxmert.py](https://codecov.io/gh/huggingface/transformers/pull/5793/diff?src=pr&el=tree#diff-c3JjL3RyYW5zZm9ybWVycy9jb25maWd1cmF0aW9uX2x4bWVydC5weQ==) | `100.00% <100.00%> (ø)` | |\n| [src/transformers/modeling\\_auto.py](https://codecov.io/gh/huggingface/transformers/pull/5793/diff?src=pr&el=tree#diff-c3JjL3RyYW5zZm9ybWVycy9tb2RlbGluZ19hdXRvLnB5) | `78.85% <100.00%> (+0.12%)` | :arrow_up: |\n| [src/transformers/tokenization\\_auto.py](https://codecov.io/gh/huggingface/transformers/pull/5793/diff?src=pr&el=tree#diff-c3JjL3RyYW5zZm9ybWVycy90b2tlbml6YXRpb25fYXV0by5weQ==) | `97.82% <100.00%> (+2.27%)` | :arrow_up: |\n| [src/transformers/tokenization\\_lxmert.py](https://codecov.io/gh/huggingface/transformers/pull/5793/diff?src=pr&el=tree#diff-c3JjL3RyYW5zZm9ybWVycy90b2tlbml6YXRpb25fbHhtZXJ0LnB5) | `100.00% <100.00%> (ø)` | |\n| [src/transformers/modeling\\_reformer.py](https://codecov.io/gh/huggingface/transformers/pull/5793/diff?src=pr&el=tree#diff-c3JjL3RyYW5zZm9ybWVycy9tb2RlbGluZ19yZWZvcm1lci5weQ==) | `16.87% <0.00%> (-79.30%)` | :arrow_down: |\n| ... and [23 more](https://codecov.io/gh/huggingface/transformers/pull/5793/diff?src=pr&el=tree-more) | |\n\n------\n\n[Continue to review full report at Codecov](https://codecov.io/gh/huggingface/transformers/pull/5793?src=pr&el=continue).\n> **Legend** - [Click here to learn more](https://docs.codecov.io/docs/codecov-delta)\n> `Δ = absolute <relative> (impact)`, `ø = not affected`, `? = missing data`\n> Powered by [Codecov](https://codecov.io/gh/huggingface/transformers/pull/5793?src=pr&el=footer). Last update [930153e...4ed21b4](https://codecov.io/gh/huggingface/transformers/pull/5793?src=pr&el=lastupdated). Read the [comment docs](https://docs.codecov.io/docs/pull-request-comments).\n" ]

[]

[ { "login": "sshleifer", "id": 6045025, "node_id": "MDQ6VXNlcjYwNDUwMjU=", "avatar_url": "https://avatars.githubusercontent.com/u/6045025?v=4", "gravatar_id": "", "url": "https://api.github.com/users/sshleifer", "html_url": "https://github.com/sshleifer", "followers_url": "https://api.github.com/users/sshleifer/followers", "following_url": "https://api.github.com/users/sshleifer/following{/other_user}", "gists_url": "https://api.github.com/users/sshleifer/gists{/gist_id}", "starred_url": "https://api.github.com/users/sshleifer/starred{/owner}{/repo}", "subscriptions_url": "https://api.github.com/users/sshleifer/subscriptions", "organizations_url": "https://api.github.com/users/sshleifer/orgs", "repos_url": "https://api.github.com/users/sshleifer/repos", "events_url": "https://api.github.com/users/sshleifer/events{/privacy}", "received_events_url": "https://api.github.com/users/sshleifer/received_events", "type": "User", "site_admin": false } ]

[ "The isort check doesn't fail when I run it on my end. Other than that - I would say this is complete.\r\n", "Awesome, I'm testing it now on WMT english-romanian translation.", "OK, I have it running with many modifications and it works much better. Very little cpu ram wasted!\r\n\r\nIs it OK if I make a new PR or do you prefer to add me as a contributor to your fork and so I can push to this branch? either way you will get all the credit in the release notes/pr summary :) \r\n\r\n", "Thanks - I'll add you as a contributor in mine. :)", "> Thanks - I'll add you as a contributor in mine. :)\r\n\r\nYou can just click on \"Allow edits from maintainers\" on your PR. (in case you didn't know this feature)", "Any updates here? Is it good to go?\r\n", "I'm still working on cleaning up my code. Sorry for the delay. ", "Biggest change:\r\n- For MBart Tokenizer, we can't use the `encode_line` approach because there are special tokens all over the place, so I made a separate dataset. \r\n\r\nStylistic:\r\n- Address the `linecache` off by 1 error inside of `__getitem__` instead other places.\r\n- `_get_examples` -> `get_char_lens`.\r\n- `MbartTokenizer` cleanup.", "# [Codecov](https://codecov.io/gh/huggingface/transformers/pull/5792?src=pr&el=h1) Report\n> Merging [#5792](https://codecov.io/gh/huggingface/transformers/pull/5792?src=pr&el=desc) into [master](https://codecov.io/gh/huggingface/transformers/commit/eae6d8d14f1d25d62c3fe9e7e410607bbaf69787&el=desc) will **increase** coverage by `0.91%`.\n> The diff coverage is `100.00%`.\n\n[](https://codecov.io/gh/huggingface/transformers/pull/5792?src=pr&el=tree)\n\n```diff\n@@ Coverage Diff @@\n## master #5792 +/- ##\n==========================================\n+ Coverage 77.54% 78.46% +0.91% \n==========================================\n Files 146 146 \n Lines 26200 26200 \n==========================================\n+ Hits 20318 20559 +241 \n+ Misses 5882 5641 -241 \n```\n\n\n| [Impacted Files](https://codecov.io/gh/huggingface/transformers/pull/5792?src=pr&el=tree) | Coverage Δ | |\n|---|---|---|\n| [src/transformers/tokenization\\_bart.py](https://codecov.io/gh/huggingface/transformers/pull/5792/diff?src=pr&el=tree#diff-c3JjL3RyYW5zZm9ybWVycy90b2tlbml6YXRpb25fYmFydC5weQ==) | `95.45% <100.00%> (ø)` | |\n| [src/transformers/modeling\\_tf\\_electra.py](https://codecov.io/gh/huggingface/transformers/pull/5792/diff?src=pr&el=tree#diff-c3JjL3RyYW5zZm9ybWVycy9tb2RlbGluZ190Zl9lbGVjdHJhLnB5) | `26.02% <0.00%> (-69.52%)` | :arrow_down: |\n| [src/transformers/generation\\_tf\\_utils.py](https://codecov.io/gh/huggingface/transformers/pull/5792/diff?src=pr&el=tree#diff-c3JjL3RyYW5zZm9ybWVycy9nZW5lcmF0aW9uX3RmX3V0aWxzLnB5) | `85.21% <0.00%> (-1.26%)` | :arrow_down: |\n| [src/transformers/modeling\\_tf\\_mobilebert.py](https://codecov.io/gh/huggingface/transformers/pull/5792/diff?src=pr&el=tree#diff-c3JjL3RyYW5zZm9ybWVycy9tb2RlbGluZ190Zl9tb2JpbGViZXJ0LnB5) | `96.77% <0.00%> (+73.38%)` | :arrow_up: |\n\n------\n\n[Continue to review full report at Codecov](https://codecov.io/gh/huggingface/transformers/pull/5792?src=pr&el=continue).\n> **Legend** - [Click here to learn more](https://docs.codecov.io/docs/codecov-delta)\n> `Δ = absolute <relative> (impact)`, `ø = not affected`, `? = missing data`\n> Powered by [Codecov](https://codecov.io/gh/huggingface/transformers/pull/5792?src=pr&el=footer). Last update [eae6d8d...79d73ee](https://codecov.io/gh/huggingface/transformers/pull/5792?src=pr&el=lastupdated). Read the [comment docs](https://docs.codecov.io/docs/pull-request-comments).\n" ]

[]

[ { "login": "LysandreJik", "id": 30755778, "node_id": "MDQ6VXNlcjMwNzU1Nzc4", "avatar_url": "https://avatars.githubusercontent.com/u/30755778?v=4", "gravatar_id": "", "url": "https://api.github.com/users/LysandreJik", "html_url": "https://github.com/LysandreJik", "followers_url": "https://api.github.com/users/LysandreJik/followers", "following_url": "https://api.github.com/users/LysandreJik/following{/other_user}", "gists_url": "https://api.github.com/users/LysandreJik/gists{/gist_id}", "starred_url": "https://api.github.com/users/LysandreJik/starred{/owner}{/repo}", "subscriptions_url": "https://api.github.com/users/LysandreJik/subscriptions", "organizations_url": "https://api.github.com/users/LysandreJik/orgs", "repos_url": "https://api.github.com/users/LysandreJik/repos", "events_url": "https://api.github.com/users/LysandreJik/events{/privacy}", "received_events_url": "https://api.github.com/users/LysandreJik/received_events", "type": "User", "site_admin": false } ]

[ "# [Codecov](https://codecov.io/gh/huggingface/transformers/pull/5791?src=pr&el=h1) Report\n> Merging [#5791](https://codecov.io/gh/huggingface/transformers/pull/5791?src=pr&el=desc) into [master](https://codecov.io/gh/huggingface/transformers/commit/3b924fabeef717be8399f1888280c29c69e9ab00&el=desc) will **decrease** coverage by `0.08%`.\n> The diff coverage is `n/a`.\n\n[](https://codecov.io/gh/huggingface/transformers/pull/5791?src=pr&el=tree)\n\n```diff\n@@ Coverage Diff @@\n## master #5791 +/- ##\n==========================================\n- Coverage 78.13% 78.05% -0.09% \n==========================================\n Files 146 146 \n Lines 26047 26047 \n==========================================\n- Hits 20352 20330 -22 \n- Misses 5695 5717 +22 \n```\n\n\n| [Impacted Files](https://codecov.io/gh/huggingface/transformers/pull/5791?src=pr&el=tree) | Coverage Δ | |\n|---|---|---|\n| [src/transformers/modeling\\_tf\\_distilbert.py](https://codecov.io/gh/huggingface/transformers/pull/5791/diff?src=pr&el=tree#diff-c3JjL3RyYW5zZm9ybWVycy9tb2RlbGluZ190Zl9kaXN0aWxiZXJ0LnB5) | `64.90% <0.00%> (-33.90%)` | :arrow_down: |\n| [src/transformers/generation\\_tf\\_utils.py](https://codecov.io/gh/huggingface/transformers/pull/5791/diff?src=pr&el=tree#diff-c3JjL3RyYW5zZm9ybWVycy9nZW5lcmF0aW9uX3RmX3V0aWxzLnB5) | `85.96% <0.00%> (-0.26%)` | :arrow_down: |\n| [src/transformers/generation\\_utils.py](https://codecov.io/gh/huggingface/transformers/pull/5791/diff?src=pr&el=tree#diff-c3JjL3RyYW5zZm9ybWVycy9nZW5lcmF0aW9uX3V0aWxzLnB5) | `97.11% <0.00%> (+0.28%)` | :arrow_up: |\n| [src/transformers/modeling\\_tf\\_roberta.py](https://codecov.io/gh/huggingface/transformers/pull/5791/diff?src=pr&el=tree#diff-c3JjL3RyYW5zZm9ybWVycy9tb2RlbGluZ190Zl9yb2JlcnRhLnB5) | `93.36% <0.00%> (+49.37%)` | :arrow_up: |\n\n------\n\n[Continue to review full report at Codecov](https://codecov.io/gh/huggingface/transformers/pull/5791?src=pr&el=continue).\n> **Legend** - [Click here to learn more](https://docs.codecov.io/docs/codecov-delta)\n> `Δ = absolute <relative> (impact)`, `ø = not affected`, `? = missing data`\n> Powered by [Codecov](https://codecov.io/gh/huggingface/transformers/pull/5791?src=pr&el=footer). Last update [3b924fa...378b034](https://codecov.io/gh/huggingface/transformers/pull/5791?src=pr&el=lastupdated). Read the [comment docs](https://docs.codecov.io/docs/pull-request-comments).\n", "Other than that, LGTM! Thanks for your work!", "Great, just renamed it - let me know if anything else should be changed!" ]

[]

[]

[ "# [Codecov](https://codecov.io/gh/huggingface/transformers/pull/5790?src=pr&el=h1) Report\n> Merging [#5790](https://codecov.io/gh/huggingface/transformers/pull/5790?src=pr&el=desc) into [master](https://codecov.io/gh/huggingface/transformers/commit/d088d744adb4e5aa45262a34acab3ae9e81de169&el=desc) will **decrease** coverage by `0.84%`.\n> The diff coverage is `n/a`.\n\n[](https://codecov.io/gh/huggingface/transformers/pull/5790?src=pr&el=tree)\n\n```diff\n@@ Coverage Diff @@\n## master #5790 +/- ##\n==========================================\n- Coverage 78.10% 77.26% -0.85% \n==========================================\n Files 146 146 \n Lines 26047 26047 \n==========================================\n- Hits 20344 20125 -219 \n- Misses 5703 5922 +219 \n```\n\n\n| [Impacted Files](https://codecov.io/gh/huggingface/transformers/pull/5790?src=pr&el=tree) | Coverage Δ | |\n|---|---|---|\n| [src/transformers/modeling\\_tf\\_mobilebert.py](https://codecov.io/gh/huggingface/transformers/pull/5790/diff?src=pr&el=tree#diff-c3JjL3RyYW5zZm9ybWVycy9tb2RlbGluZ190Zl9tb2JpbGViZXJ0LnB5) | `23.38% <0.00%> (-73.39%)` | :arrow_down: |\n| [src/transformers/generation\\_tf\\_utils.py](https://codecov.io/gh/huggingface/transformers/pull/5790/diff?src=pr&el=tree#diff-c3JjL3RyYW5zZm9ybWVycy9nZW5lcmF0aW9uX3RmX3V0aWxzLnB5) | `83.95% <0.00%> (-2.01%)` | :arrow_down: |\n| [src/transformers/modeling\\_openai.py](https://codecov.io/gh/huggingface/transformers/pull/5790/diff?src=pr&el=tree#diff-c3JjL3RyYW5zZm9ybWVycy9tb2RlbGluZ19vcGVuYWkucHk=) | `82.31% <0.00%> (+1.28%)` | :arrow_up: |\n| [src/transformers/modeling\\_tf\\_openai.py](https://codecov.io/gh/huggingface/transformers/pull/5790/diff?src=pr&el=tree#diff-c3JjL3RyYW5zZm9ybWVycy9tb2RlbGluZ190Zl9vcGVuYWkucHk=) | `95.18% <0.00%> (+74.91%)` | :arrow_up: |\n\n------\n\n[Continue to review full report at Codecov](https://codecov.io/gh/huggingface/transformers/pull/5790?src=pr&el=continue).\n> **Legend** - [Click here to learn more](https://docs.codecov.io/docs/codecov-delta)\n> `Δ = absolute <relative> (impact)`, `ø = not affected`, `? = missing data`\n> Powered by [Codecov](https://codecov.io/gh/huggingface/transformers/pull/5790?src=pr&el=footer). Last update [d088d74...46f7d0f](https://codecov.io/gh/huggingface/transformers/pull/5790?src=pr&el=lastupdated). Read the [comment docs](https://docs.codecov.io/docs/pull-request-comments).\n", "I like it! Maybe we can also put in **big** that people should add their env information and not insert screenshots - this happens still quite often from what I see", "@sshleifer I think I've committed instead of suggesting an edit as I wasn't a reviewer, sorry, lemme know if it broke anything!", "ok to merge @julien-c ?", "Like this too! Added myself for issues linked to the documentation, feel free to add more my way.", "@sshleifer could you tag me for `examples/token-classification` 🤔" ]

[ { "id": 1838412367, "node_id": "MDU6TGFiZWwxODM4NDEyMzY3", "url": "https://api.github.com/repos/huggingface/transformers/labels/model%20card", "name": "model card", "color": "92d5f4", "default": false, "description": "Related to pretrained model cards" } ]

[]

[ "# [Codecov](https://codecov.io/gh/huggingface/transformers/pull/5789?src=pr&el=h1) Report\n> Merging [#5789](https://codecov.io/gh/huggingface/transformers/pull/5789?src=pr&el=desc) into [master](https://codecov.io/gh/huggingface/transformers/commit/3b924fabeef717be8399f1888280c29c69e9ab00&el=desc) will **increase** coverage by `0.11%`.\n> The diff coverage is `n/a`.\n\n[](https://codecov.io/gh/huggingface/transformers/pull/5789?src=pr&el=tree)\n\n```diff\n@@ Coverage Diff @@\n## master #5789 +/- ##\n==========================================\n+ Coverage 78.13% 78.25% +0.11% \n==========================================\n Files 146 146 \n Lines 26047 26047 \n==========================================\n+ Hits 20352 20383 +31 \n+ Misses 5695 5664 -31 \n```\n\n\n| [Impacted Files](https://codecov.io/gh/huggingface/transformers/pull/5789?src=pr&el=tree) | Coverage Δ | |\n|---|---|---|\n| [src/transformers/modeling\\_tf\\_electra.py](https://codecov.io/gh/huggingface/transformers/pull/5789/diff?src=pr&el=tree#diff-c3JjL3RyYW5zZm9ybWVycy9tb2RlbGluZ190Zl9lbGVjdHJhLnB5) | `26.02% <0.00%> (-69.52%)` | :arrow_down: |\n| [src/transformers/generation\\_tf\\_utils.py](https://codecov.io/gh/huggingface/transformers/pull/5789/diff?src=pr&el=tree#diff-c3JjL3RyYW5zZm9ybWVycy9nZW5lcmF0aW9uX3RmX3V0aWxzLnB5) | `85.21% <0.00%> (-1.01%)` | :arrow_down: |\n| [src/transformers/modeling\\_openai.py](https://codecov.io/gh/huggingface/transformers/pull/5789/diff?src=pr&el=tree#diff-c3JjL3RyYW5zZm9ybWVycy9tb2RlbGluZ19vcGVuYWkucHk=) | `82.31% <0.00%> (+1.28%)` | :arrow_up: |\n| [src/transformers/modeling\\_tf\\_openai.py](https://codecov.io/gh/huggingface/transformers/pull/5789/diff?src=pr&el=tree#diff-c3JjL3RyYW5zZm9ybWVycy9tb2RlbGluZ190Zl9vcGVuYWkucHk=) | `95.18% <0.00%> (+74.91%)` | :arrow_up: |\n\n------\n\n[Continue to review full report at Codecov](https://codecov.io/gh/huggingface/transformers/pull/5789?src=pr&el=continue).\n> **Legend** - [Click here to learn more](https://docs.codecov.io/docs/codecov-delta)\n> `Δ = absolute <relative> (impact)`, `ø = not affected`, `? = missing data`\n> Powered by [Codecov](https://codecov.io/gh/huggingface/transformers/pull/5789?src=pr&el=footer). Last update [3b924fa...f284d8c](https://codecov.io/gh/huggingface/transformers/pull/5789?src=pr&el=lastupdated). Read the [comment docs](https://docs.codecov.io/docs/pull-request-comments).\n" ]

[ { "id": 2206883508, "node_id": "MDU6TGFiZWwyMjA2ODgzNTA4", "url": "https://api.github.com/repos/huggingface/transformers/labels/lightning", "name": "lightning", "color": "a707bc", "default": false, "description": "" } ]

[]

[ "This is a duplicate." ]

[ { "id": 1314768611, "node_id": "MDU6TGFiZWwxMzE0NzY4NjEx", "url": "https://api.github.com/repos/huggingface/transformers/labels/wontfix", "name": "wontfix", "color": "ffffff", "default": true, "description": null } ]

[]

[ "Hi! This is probably because the error reponse for a request is silenced . You can place a breakpont [here](https://github.com/huggingface/transformers/blob/0533cf470659b97c6279bd04f65536a1ec88404a/src/transformers/file_utils.py#L681) and check. Mine was SSL error, so I set `REQUESTS_CA_BUNDLE` env var to `/etc/ssl/certs/ca-certificates.crt`", "@festeh Which variable should I be looking at in the breakpoint? If it's `response`, which attribute?", "You can just type `requests.head(url, allow_redirects=True, proxies=proxies, timeout=etag_timeout)` in debugger console and check exception traceback.", "I gave that a shot and got the following:\r\n```python\r\nIn[7]: requests.head(url, allow_redirects=True, proxies=proxies, timeout=etag_timeout)\r\nOut[7]: <Response [200]>\r\n```", "Interesting, probably you have an error around this request then, when you're actually downloading weights\r\nhttps://github.com/huggingface/transformers/blob/7fad617dc1fc681a7f5da5e0172c8b83f4bf0024/src/transformers/file_utils.py#L678\r\n\r\nif you place a breakpoint here, would the program hit it? and if not could you send this request manually?", "I think the program hits it:\r\n```\r\nrequests.get(url, stream=True, proxies=proxies, headers=headers)\r\nOut[2]: <Response [200]>\r\n```", "Well, if you get the message from the first post it means that some line of code has raised an `EnvironmentError` or `TimeoutError`. I think you need to advance over all lines in `get_from_cache` and find out which line is responsible for that. After you find this line, you can re-run it in console and see the actual exception.", "Ok here is what I found out. If I place a breakpoint at `etag = None` in `get_from_cache()` in `file_utils.py` and try to run things, it stops by that breakpoint twice. The first time, I step through to `response = requests.head(url, allow_redirects=True, proxies=proxies, timeout=etag_timeout)` and get the following response (with the inputs shown from the debugger):\r\n```python\r\ncache_dir = {str} '/path/to/.cache/torch/transformers'\r\netag = {NoneType} None\r\netag_timeout = {int} 10\r\nforce_download = {bool} False\r\nlocal_files_only = {bool} False\r\nproxies = {dict: 3} {'http': 'http://myproxy.com:port', 'https': 'https://myproxy.com:port', 'no': ',127.0.0.1,127.0.0.111,127.0.0.2'}\r\nresponse = {Response} <Response [200]>\r\nresume_download = {bool} False\r\nurl = {str} 'https://s3.amazonaws.com/models.huggingface.co/bert/gpt2-config.json'\r\nuser_agent = {NoneType} None\r\n\r\nrequests.head(url, allow_redirects=True, proxies=proxies, timeout=etag_timeout)\r\nOut[1]: <Response [200]>\r\n```\r\n\r\non the second time, I get:\r\n```python\r\ncache_dir = {str} '/path/to/.cache/torch/transformers'\r\netag = {NoneType} None\r\netag_timeout = {int} 10\r\nforce_download = {bool} False\r\nlocal_files_only = {bool} False\r\nproxies = {dict: 3} {'http': 'http://myproxy.com:port', 'https': 'https://myproxy.com:port', 'no': ',127.0.0.1,127.0.0.111,127.0.0.2'}\r\nresume_download = {bool} False\r\nurl = {str} 'https://cdn.huggingface.co/gpt2-pytorch_model.bin'\r\nuser_agent = {NoneType} None\r\n\r\nrequests.head(url, allow_redirects=True, proxies=proxies, timeout=etag_timeout)\r\nTraceback (most recent call last):\r\n File \"/path/to/anaconda3/lib/python3.7/site-packages/urllib3/contrib/pyopenssl.py\", line 485, in wrap_socket\r\n cnx.do_handshake()\r\n File \"/path/to/anaconda3/lib/python3.7/site-packages/OpenSSL/SSL.py\", line 1934, in do_handshake\r\n self._raise_ssl_error(self._ssl, result)\r\n File \"/path/to/anaconda3/lib/python3.7/site-packages/OpenSSL/SSL.py\", line 1671, in _raise_ssl_error\r\n _raise_current_error()\r\n File \"/path/to/anaconda3/lib/python3.7/site-packages/OpenSSL/_util.py\", line 54, in exception_from_error_queue\r\n raise exception_type(errors)\r\nOpenSSL.SSL.Error: [('SSL routines', 'tls_process_server_certificate', 'certificate verify failed')]\r\nDuring handling of the above exception, another exception occurred:\r\nTraceback (most recent call last):\r\n File \"/path/to/anaconda3/lib/python3.7/site-packages/urllib3/connectionpool.py\", line 662, in urlopen\r\n self._prepare_proxy(conn)\r\n File \"/path/to/anaconda3/lib/python3.7/site-packages/urllib3/connectionpool.py\", line 948, in _prepare_proxy\r\n conn.connect()\r\n File \"/path/to/anaconda3/lib/python3.7/site-packages/urllib3/connection.py\", line 360, in connect\r\n ssl_context=context,\r\n File \"/path/to/anaconda3/lib/python3.7/site-packages/urllib3/util/ssl_.py\", line 370, in ssl_wrap_socket\r\n return context.wrap_socket(sock, server_hostname=server_hostname)\r\n File \"/path/to/anaconda3/lib/python3.7/site-packages/urllib3/contrib/pyopenssl.py\", line 491, in wrap_socket\r\n raise ssl.SSLError(\"bad handshake: %r\" % e)\r\nssl.SSLError: (\"bad handshake: Error([('SSL routines', 'tls_process_server_certificate', 'certificate verify failed')])\",)\r\nDuring handling of the above exception, another exception occurred:\r\nTraceback (most recent call last):\r\n File \"/path/to/anaconda3/lib/python3.7/site-packages/requests/adapters.py\", line 449, in send\r\n timeout=timeout\r\n File \"/path/to/anaconda3/lib/python3.7/site-packages/urllib3/connectionpool.py\", line 720, in urlopen\r\n method, url, error=e, _pool=self, _stacktrace=sys.exc_info()[2]\r\n File \"/path/to/anaconda3/lib/python3.7/site-packages/urllib3/util/retry.py\", line 436, in increment\r\n raise MaxRetryError(_pool, url, error or ResponseError(cause))\r\nurllib3.exceptions.MaxRetryError: HTTPSConnectionPool(host='cdn.huggingface.co', port=443): Max retries exceeded with url: /gpt2-pytorch_model.bin (Caused by SSLError(SSLError(\"bad handshake: Error([('SSL routines', 'tls_process_server_certificate', 'certificate verify failed')])\")))\r\nDuring handling of the above exception, another exception occurred:\r\nTraceback (most recent call last):\r\n File \"/path/to/anaconda3/lib/python3.7/site-packages/IPython/core/interactiveshell.py\", line 3331, in run_code\r\n exec(code_obj, self.user_global_ns, self.user_ns)\r\n File \"<ipython-input-3-5b27aae00c67>\", line 1, in <module>\r\n requests.head(url, allow_redirects=True, proxies=proxies, timeout=etag_timeout)\r\n File \"/path/to/anaconda3/lib/python3.7/site-packages/requests/api.py\", line 101, in head\r\n return request('head', url, **kwargs)\r\n File \"/path/to/anaconda3/lib/python3.7/site-packages/requests/api.py\", line 60, in request\r\n return session.request(method=method, url=url, **kwargs)\r\n File \"/path/to/anaconda3/lib/python3.7/site-packages/requests/sessions.py\", line 533, in request\r\n resp = self.send(prep, **send_kwargs)\r\n File \"/path/to/anaconda3/lib/python3.7/site-packages/requests/sessions.py\", line 646, in send\r\n r = adapter.send(request, **kwargs)\r\n File \"/path/to/anaconda3/lib/python3.7/site-packages/requests/adapters.py\", line 514, in send\r\n raise SSLError(e, request=request)\r\nrequests.exceptions.SSLError: HTTPSConnectionPool(host='cdn.huggingface.co', port=443): Max retries exceeded with url: /gpt2-pytorch_model.bin (Caused by SSLError(SSLError(\"bad handshake: Error([('SSL routines', 'tls_process_server_certificate', 'certificate verify failed')])\")))\r\n```\r\nAlso on the second time, it skips right over the `if response.status_code == 200:` check and goes straight to `except (EnvironmentError, requests.exceptions.Timeout):` when I advance a step from `response = requests.head(url, allow_redirects=True, proxies=proxies, timeout=etag_timeout)`\r\n```python\r\netag = None\r\n if not local_files_only:\r\n try:\r\n response = requests.head(url, allow_redirects=True, proxies=proxies, timeout=etag_timeout)\r\n if response.status_code == 200:\r\n etag = response.headers.get(\"ETag\")\r\n except (EnvironmentError, requests.exceptions.Timeout):\r\n # etag is already None\r\n pass\r\n```", "Any further thoughts on this?", "I'm suffering same problem. Cannot use behind the proxy servers (private network).", "This issue has been automatically marked as stale because it has not had recent activity. It will be closed if no further activity occurs. Thank you for your contributions.\n", "I have the same problem. `AutoModelForSeq2SeqLM.from_pretrained(\"google/pegasus-xsum\")` just idles forever.\r\nLast version of transformers and torch. \r\nIm not using proxy, but if I use vpn 'from_pretrained' is actually downloading the model", "This issue has been automatically marked as stale because it has not had recent activity. It will be closed if no further activity occurs. Thank you for your contributions.\n", "This issue has been automatically marked as stale and been closed because it has not had recent activity. Thank you for your contributions.\n\nIf you think this still needs to be addressed please comment on this thread." ]

[ { "id": 1108649053, "node_id": "MDU6TGFiZWwxMTA4NjQ5MDUz", "url": "https://api.github.com/repos/huggingface/transformers/labels/Help%20wanted", "name": "Help wanted", "color": "008672", "default": false, "description": "Extra attention is needed, help appreciated" }, { "id": 1845609017, "node_id": "MDU6TGFiZWwxODQ1NjA5MDE3", "url": "https://api.github.com/repos/huggingface/transformers/labels/seq2seq", "name": "seq2seq", "color": "fef2c0", "default": false, "description": "" }, { "id": 1936351150, "node_id": "MDU6TGFiZWwxOTM2MzUxMTUw", "url": "https://api.github.com/repos/huggingface/transformers/labels/Examples", "name": "Examples", "color": "d4c5f9", "default": false, "description": "Which is related to examples in general" } ]

[ { "login": "sshleifer", "id": 6045025, "node_id": "MDQ6VXNlcjYwNDUwMjU=", "avatar_url": "https://avatars.githubusercontent.com/u/6045025?v=4", "gravatar_id": "", "url": "https://api.github.com/users/sshleifer", "html_url": "https://github.com/sshleifer", "followers_url": "https://api.github.com/users/sshleifer/followers", "following_url": "https://api.github.com/users/sshleifer/following{/other_user}", "gists_url": "https://api.github.com/users/sshleifer/gists{/gist_id}", "starred_url": "https://api.github.com/users/sshleifer/starred{/owner}{/repo}", "subscriptions_url": "https://api.github.com/users/sshleifer/subscriptions", "organizations_url": "https://api.github.com/users/sshleifer/orgs", "repos_url": "https://api.github.com/users/sshleifer/repos", "events_url": "https://api.github.com/users/sshleifer/events{/privacy}", "received_events_url": "https://api.github.com/users/sshleifer/received_events", "type": "User", "site_admin": false } ]

[ "Hi @sshleifer , can embed pruning affect accuracy(bleu) ? TPU option seems good as it won't result in less accuracy ", "https://github.com/pytorch/fairseq/issues/2120#issuecomment-647915216 suggests it costs 2 BLEU points.\r\n\r\nOne of our goals with examples is to make train loops that people can run on their hardware+data. If we can get 36 on a 16GB gpu that is super useful, just like `--freeze_encoder` and `--freeze_embeds` are useful, even if they probably hurt final performance a bit.\r\n", "Broke into smaller issues, but leaving this open in case people have other ideas!", "note: fairseq wmt_en_de batch wps=14721.8 on V100 with https://github.com/pytorch/fairseq/issues/2506#issuecomment-678630596\r\n\r\ncheck tpb/seconds for seq2seq/finetune.py\r\n\r\n", "I am trying to fine-tune on wmt-en-ro (facebook/mbart-large-cc25) using Colab and Kaggle 16GB gpu. On Colab, I get cuda oom and on Kaggle I get out of disk apace (20 GB limit). Is there a way I can skip checkpoints to fit the process in 20GB disk space? \r\nI even tried 1 bs, 8 max len, fp16 and freeze_encoder.", "High level, I would use an opus-mt model instead of mbart for most tasks. They are smaller and tend to be nearly as good at translation, if not better. I have run ~50~ experiments finetuning mbart on wmt-en-ro and it was not a particularly pleasant experience.\r\n\r\nDisk space: You can see what happens if you remove `checkpoint_callback` from \r\nhttps://github.com/huggingface/transformers/blob/9336086ab5d232cccd9512333518cf4299528882/examples/seq2seq/finetune.py#L362\r\n\r\nand just call `model.model.save_pretrained` `model.model.half().save_pretrained` after.\r\n\r\n\r\n\r\n\r\n", "Thanks for the advice @sshleifer - just wanted to try it out and test other language pairs. \r\nOn Kaggle, as I just found out, the trick is to create the output_dir outside the working directory (where total disk space is just 5GB). Kaggle won't save it with kernel commit though.\r\n", "Cool! LMK if you have good results!", "I got it working using a small subset of data, then went on training English -> Arabic translator (just a proof of concept). Model uploaded to [HF](https://huggingface.co/akhooli/mbart-large-cc25-en-ar).", "Hi, I had written the code to do vocab pruning and because of the number of people wanting help with it I converted the code into a standalone library. I hope its okay if I link it here for other people to find. \r\n\r\n[Link to repo.](https://github.com/IamAdiSri/hf-trim)\r\n\r\nReferencing issues #5896 #6132" ]

[]

[]

[ "This is just one of those days.. There's a[ mean pooling function](https://huggingface.co/sentence-transformers/bert-base-nli-mean-tokens) here that can be adapted." ]

[]

[]

[]

[ { "id": 2039044877, "node_id": "MDU6TGFiZWwyMDM5MDQ0ODc3", "url": "https://api.github.com/repos/huggingface/transformers/labels/marian", "name": "marian", "color": "30cc95", "default": false, "description": "" } ]

[ { "login": "sshleifer", "id": 6045025, "node_id": "MDQ6VXNlcjYwNDUwMjU=", "avatar_url": "https://avatars.githubusercontent.com/u/6045025?v=4", "gravatar_id": "", "url": "https://api.github.com/users/sshleifer", "html_url": "https://github.com/sshleifer", "followers_url": "https://api.github.com/users/sshleifer/followers", "following_url": "https://api.github.com/users/sshleifer/following{/other_user}", "gists_url": "https://api.github.com/users/sshleifer/gists{/gist_id}", "starred_url": "https://api.github.com/users/sshleifer/starred{/owner}{/repo}", "subscriptions_url": "https://api.github.com/users/sshleifer/subscriptions", "organizations_url": "https://api.github.com/users/sshleifer/orgs", "repos_url": "https://api.github.com/users/sshleifer/repos", "events_url": "https://api.github.com/users/sshleifer/events{/privacy}", "received_events_url": "https://api.github.com/users/sshleifer/received_events", "type": "User", "site_admin": false } ]

[ "The script you want is at `src/transformers/convert_marian_to_pytorch.py` \r\nIt requires you to download the marian model you wish to convert and also to clone\r\n```bash\r\ngit clone [email protected]:Helsinki-NLP/Opus-MT-train.git\r\n```\r\nyou may have to adjust some paths in the script (like `repo_path`) based on where things are.\r\nhttps://github.com/huggingface/transformers/blob/448c467256332e4be8c122a159b482c1ef039b98/src/transformers/convert_marian_to_pytorch.py#L189\r\n", "Also note that we have ported 1000+ of them, and some were renamed. Which one are you looking for?", "I have trained a Transformer model to translate from Italian to Dutch with Marian based on https://github.com/marian-nmt/marian-examples/tree/master/transformer and data from OPUS. The model is using BPE for tokenization. I have the itnl.bpe, vocab file, model.npz file, etc. on my computer. \r\n\r\nThe example with the conver_marian_to_pytorch.py is using the model files from the Helsinki-NLP/Opus-MT-train repo instead of local model files. So how can I use the local model files (that I trained with marian) to convert them to a pytorch model that can be used with huggingface?\r\n\r\n@sshleifer ", "I don't think BPE tokenizer will work. To answer the local model question, you need to (roughly) (a) run the converter on a model from the repo and see what files get downloaded, (b) make your filesystem look like that (c) update the code to not download things and not make model cards.", "I am probably missing something here. I am trying to buid a grammar corrector using Marian NMT generated model with Huggingface's transformer. Source language is the text with erros and target language the text without. I have trained the model following this example (https://github.com/marian-nmt/marian-examples/tree/master/transformer). As it does not generate source and target spm files, I created both of them them using \"build/spm_train\" provided with Marian implementation and, of course, using for each one their respective training files, the same used for training the model.\r\n\r\nThe commands to generate spm files are:\r\n ../../build/spm_train --input data/src_sentences_dev.txt --model_prefix=source --vocab_size=16000 --character_coverage=1.0\r\n../../build/spm_train --input data/ref_sentences_dev.txt --model_prefix=target --vocab_size=16000 --character_coverage=1.0\r\n\r\nAfter that I proceeded with the convertion to pythorch using https://github.com/huggingface/transformers/blob/master/src/transformers/models/marian/convert_marian_to_pytorch.py. This convertion went fine. The problem is that when I use this model in Huggingface's transformer.MarianMTModel and MarianTokenizer (from_pretrained) I get weired results.\r\n\r\nSo to mitigate that I tried to perform another convertion, this time using en-de files downloaded from https://github.com/Helsinki-NLP/OPUS-MT-train/tree/master/models/en-de. Convertion went fine, unfortunately with the same weired results, example at the bottom, as with the model I created for grammar correction.\r\n\r\nI further downloaded the required files from https://huggingface.co/Helsinki-NLP/opus-mt-en-de/tree/main and substituted the ones genereated by convert_marian_to_pytorch.py above with these and, as expected, the translation went fine.\r\n\r\nThe converter, to my knowledge, requires 4 files: source.spm, target.spm, vocab.yml and model.npz. As these weired results raises after I use the converter, both with my model and the en-de downloaded from https://github.com/Helsinki-NLP/OPUS-MT-train/tree/master/models/en-de, I guess I am missing a piece of info that I cannot identify.\r\n\r\nIt is worth noting that my grammar corrector model works correctly with marian_decoder.\r\n\r\nAny help will be very much appreciated! @sshleifer ?\r\n\r\nCheers\r\n\r\nExample of English -> German translation:\r\nSource: \"Today is Sunday\"\r\nTranslation: \r\n▁Rumänien▁Rumänien▁Rumänien▁Rumänien▁Rumänien▁Rumänien▁Rumänien▁Rumänien▁Rumänien▁Rumänien▁Rumänien▁Rumänien▁Rumänien▁Rumänien▁Rumänien▁Rumänien▁Rumänien▁Rumänien▁Rumänien▁Rumänien▁Rumänien▁Nachricht▁Rumänien▁Rumänien▁Rumänien▁Rumänien▁Rumänien▁Rumänien▁Rumänien▁Rumänien▁Rumänien▁Rumänien▁Rumänien▁Rumänien▁Rumänien▁Rumänien▁Rumänien▁Rumänien▁Rumänien▁Rumänien▁Rumänien▁Rumänien▁Rumänien▁Rumänien▁Rumänien▁Rumänien▁Rumänien▁Rumänien▁Rumänien▁Rumänien▁Rumänien▁Rumänien▁Rumänien▁Rumänien▁Rumänien▁Rumänien▁Rumänien▁Rumänien▁Rumänien▁Rumänien▁Nachricht▁Rumänien▁Rumänien▁Nachricht▁Rumänien▁Rumänien▁Rumänien▁Rumänien▁Nachricht▁Rumänien▁Rumänien▁Rumänien▁Rumänien▁verdächtige▁verdächtige▁verdächtige▁verdächtige▁verdächtige▁verdächtige▁verdächtige▁verdächtige▁verdächtige▁verdächtige▁verdächtige▁verdächtige▁verdächtige▁verdächtige▁verdächtige▁verdächtige▁verdächtige▁verdächtige▁verdächtige▁verdächtige▁verdächtige▁verdächtige▁verdächtige▁verdächtige▁verdächtige▁verdächtige▁verdächtige▁verdächtige▁verdächtige▁verdächtige▁verdächtige▁verdächtige▁verdächtige▁verdächtige▁verdächtige▁verdächtige▁verdächtige▁verdächtige▁verdächtige▁verdächtige▁verdächtige▁verdächtige▁verdächtige▁verdächtige▁verdächtige▁verdächtige▁verdächtige▁verdächtige▁verdächtige▁verdächtige▁verdächtige▁verdächtige▁verdächtige▁verdächtige▁verdächtige▁verdächtige▁verdächtige▁verdächtige▁verdächtige▁verdächtige▁verdächtige▁verdächtige▁verdächtige▁verdächtige▁verdächtige▁verdächtige▁verdächtige▁verdächtige▁verdächtige▁verdächtige▁verdächtige▁verdächtige▁verdächtige▁verdächtige▁verdächtige▁verdächtige▁verdächtige▁verdächtige▁verdächtige▁verdächtige▁verdächtige▁verdächtige▁verdächtige▁verdächtige▁verdächtige▁verdächtige▁verdächtige▁verdächtige▁verdächtige▁verdächtige▁verdächtige▁verdächtige▁verdächtige▁verdächtige▁verdächtige▁verdächtige▁verdächtige▁verdächtige▁verdächtige▁verdächtige▁verdächtige▁verdächtige▁verdächtige▁verdächtige▁verdächtige▁verdächtige▁verdächtige▁verdächtige▁verdächtige▁verdächtige▁verdächtige▁verdächtige▁verdächtige▁verdächtige▁verdächtige▁verdächtige▁verdächtige▁verdächtige▁verdächtige▁verdächtige▁verdächtige▁verdächtige▁verdächtige▁verdächtige▁verdächtige▁verdächtige▁verdächtige▁verdächtige▁verdächtige▁verdächtige▁verdächtige▁verdächtige▁verdächtige▁verdächtige▁verdächtige▁verdächtige▁verdächtige▁verdächtige▁verdächtige▁verdächtige▁verdächtige▁verdächtige▁verdächtige▁verdächtige▁verdächtige▁verdächtige▁verdächtige▁verdächtige▁verdächtige▁verdächtige▁verdächtige▁verdächtige▁verdächtige▁verdächtige▁verdächtige▁verdächtige▁verdächtige▁verdächtige▁verdächtige▁verdächtige▁verdächtige▁verdächtige▁verdächtige▁verdächtige▁verdächtige▁verdächtige▁verdächtige▁verdächtige▁verdächtige▁verdächtige▁verdächtige▁verdächtige▁verdächtige▁verdächtige▁verdächtige▁verdächtige▁verdächtige▁verdächtige▁verdächtige▁verdächtige▁verdächtige▁verdächtige▁verdächtige▁verdächtige▁verdächtige▁verdächtige▁verdächtige▁verdächtige▁verdächtige▁verdächtige▁verdächtige▁verdächtige▁verdächtige▁verdächtige▁verdächtige▁verdächtige▁verdächtige▁verdächtige▁verdächtige▁verdächtige▁verdächtige▁verdächtige▁verdächtige▁verdächtige▁verdächtige▁verdächtige▁verdächtige▁verdächtige▁verdächtige▁verdächtige▁verdächtige▁verdächtige▁verdächtige▁verdächtige▁verdächtige▁verdächtige▁verdächtige▁verdächtige▁verdächtige▁verdächtige▁verdächtige▁verdächtige▁verdächtige▁verdächtige▁verdächtige▁verdächtige▁verdächtige▁verdächtige▁verdächtige▁verdächtige▁verdächtige▁verdächtige▁verdächtige▁verdächtige▁verdächtige▁verdächtige▁verdächtige▁verdächtige▁verdächtige▁verdächtige▁verdächtige▁verdächtige▁verdächtige▁verdächtige▁verdächtige▁verdächtige▁verdächtige▁verdächtige▁verdächtige▁verdächtige▁verdächtige▁verdächtige▁verdächtige▁verdächtige▁verdächtige▁verdächtige▁verdächtige▁verdächtige▁verdächtige▁verdächtige▁verdächtige▁verdächtige▁verdächtige▁verdächtige▁verdächtige▁verdächtige▁verdächtige▁verdächtige▁verdächtige▁verdächtige▁verdächtige▁verdächtige▁verdächtige▁verdächtige▁verdächtige▁verdächtige▁verdächtige▁verdächtige▁verdächtige▁verdächtige▁verdächtige▁verdächtige▁verdächtige▁verdächtige▁verdächtige▁verdächtige▁verdächtige▁verdächtige▁verdächtige▁verdächtige▁verdächtige▁verdächtige▁verdächtige▁verdächtige▁verdächtige▁verdächtige▁verdächtige▁verdächtige▁verdächtige▁verdächtige▁verdächtige▁verdächtige▁verdächtige▁verdächtige▁verdächtige▁verdächtige▁verdächtige▁verdächtige▁verdächtige▁verdächtige▁verdächtige▁verdächtige▁verdächtige▁verdächtige▁verdächtige▁verdächtige▁verdächtige▁verdächtige▁verdächtige▁verdächtige▁verdächtige▁verdächtige▁verdächtige▁verdächtige▁verdächtige▁verdächtige▁verdächtige▁verdächtige▁verdächtige▁verdächtige▁verdächtige▁verdächtige▁verdächtige▁verdächtige▁verdächtige▁verdächtige▁verdächtige▁verdächtige▁verdächtige▁verdächtige▁verdächtige▁verdächtige▁verdächtige▁verdächtige▁verdächtige▁verdächtige▁verdächtige▁verdächtige▁verdächtige▁verdächtige▁verdächtige▁verdächtige▁verdächtige▁verdächtige▁verdächtige▁verdächtige▁verdächtige▁verdächtige▁verdächtige▁verdächtige▁verdächtige▁verdächtige▁verdächtige▁verdächtige▁verdächtige▁verdächtige▁verdächtige▁verdächtige▁verdächtige▁verdächtige▁verdächtige▁verdächtige▁verdächtige▁verdächtige▁verdächtige▁verdächtige▁verdächtige▁verdächtige▁verdächtige▁verdächtige▁verdächtige▁verdächtige▁verdächtige▁verdächtige▁verdächtige▁verdächtige▁verdächtige▁verdächtige▁verdächtige▁verdächtige▁verdächtige▁verdächtige▁verdächtige▁verdächtige▁verdächtige▁verdächtige▁verdächtige▁verdächtige▁verdächtige▁verdächtige▁verdächtige▁verdächtige▁verdächtige▁verdächtige▁verdächtige▁verdächtige▁verdächtige▁verdächtige▁verdächtige▁verdächtige▁verdächtige▁verdächtige▁verdächtige▁verdächtige▁verdächtige▁verdächtige▁verdächtige▁verdächtige▁verdächtige▁verdächtige▁verdächtige▁verdächtige▁verdächtige▁verdächtige▁verdächtige▁verdächtige▁verdächtige▁verdächtige▁verdächtige▁verdächtige▁verdächtige▁verdächtige▁verdächtige▁verdächtige▁verdächtige▁verdächtige▁verdächtige\r\n\r\n\r\n", "I hope it's not bad practice to necro an issue but since I can't seem to link it in a new one, I'd like to add that I'm encountering th same behaviour using the script. \r\n\r\nThe conversion went fine, but the generation output is, for me : \r\n\r\n??????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????\r\n\r\nor, sometimes \r\n\r\nlinkage linkage linkage linkage linkage linkage linkage linkage linkage linkage linkage linkage linkage linkage linkage linkage linkage linkage linkage linkage linkage linkage linkage (ad lib).\r\n\r\nI should note that I have used OPUS-CAT to fine tune an OPUS-MT model, and that it performs relatively well inside the app. Also conversion with the script went on without issues. \r\n\r\nLike the OP, if I replace the model files with the original ones (in my case,Helsinki-NLP/opus-mt-zh-en), everything is fixed, so I don't think this is a pure script/params issue.\r\n\r\nThanks in advance !" ]

[ { "id": 1838412367, "node_id": "MDU6TGFiZWwxODM4NDEyMzY3", "url": "https://api.github.com/repos/huggingface/transformers/labels/model%20card", "name": "model card", "color": "92d5f4", "default": false, "description": "Related to pretrained model cards" } ]

[]

[]

[ { "id": 1838412367, "node_id": "MDU6TGFiZWwxODM4NDEyMzY3", "url": "https://api.github.com/repos/huggingface/transformers/labels/model%20card", "name": "model card", "color": "92d5f4", "default": false, "description": "Related to pretrained model cards" } ]

[]

[]

[]

[ { "login": "patrickvonplaten", "id": 23423619, "node_id": "MDQ6VXNlcjIzNDIzNjE5", "avatar_url": "https://avatars.githubusercontent.com/u/23423619?v=4", "gravatar_id": "", "url": "https://api.github.com/users/patrickvonplaten", "html_url": "https://github.com/patrickvonplaten", "followers_url": "https://api.github.com/users/patrickvonplaten/followers", "following_url": "https://api.github.com/users/patrickvonplaten/following{/other_user}", "gists_url": "https://api.github.com/users/patrickvonplaten/gists{/gist_id}", "starred_url": "https://api.github.com/users/patrickvonplaten/starred{/owner}{/repo}", "subscriptions_url": "https://api.github.com/users/patrickvonplaten/subscriptions", "organizations_url": "https://api.github.com/users/patrickvonplaten/orgs", "repos_url": "https://api.github.com/users/patrickvonplaten/repos", "events_url": "https://api.github.com/users/patrickvonplaten/events{/privacy}", "received_events_url": "https://api.github.com/users/patrickvonplaten/received_events", "type": "User", "site_admin": false } ]

[ "Hey @Alshutin, \r\n\r\nI am not able to reproduce the error. It might be because PyTorch uses a GPU and Tensorflow does not. \r\nCould you try to run your code when disabling GPU (`export CUDA_VISIBLE_DEVICES=\"\"`) and see whether the \r\nerror persists?", "Hi! I just tried it with another version of TensorFlow. With 2.2.0 it just works.", "With 2.0.0-beta1 and CUDA_VISIBLE_DEVICES=\"\" the error persists.", "Interesting - thanks for checking! \r\n\r\n\r\nDoes it crash as well for `bert-base-uncased` and TF 2.0.0?\r\n\r\nCould you run these lines to verify?\r\n\r\n```python\r\nfrom transformers import TFAutoModel, AutoTokenizer, AutoModel\r\nimport os\r\n\r\nmodel = AutoModel.from_pretrained('bert-base-uncased')\r\nos.system('mkdir bert')\r\nmodel.save_pretrained('bert')\r\nmodel = TFAutoModel.from_pretrained('bert', from_pt=True) # crashes\r\n```\r\n\r\n@thomwolf @jplu - are we gonna force TF 2.2 in `transformers` ? ", "Can you try with the 2.0.0 release and not beta? The beta was know to have a lot of issue and a lot of fixes have been applied since.\n\n@patrickvonplaten I proposed indeed to fix the TensorFlow version to 2.2, because of some welcomed features from it. But nothing has been decided yet.", "It works with 2.0.0 stable TensorFlow release. " ]

[]

[ { "login": "sshleifer", "id": 6045025, "node_id": "MDQ6VXNlcjYwNDUwMjU=", "avatar_url": "https://avatars.githubusercontent.com/u/6045025?v=4", "gravatar_id": "", "url": "https://api.github.com/users/sshleifer", "html_url": "https://github.com/sshleifer", "followers_url": "https://api.github.com/users/sshleifer/followers", "following_url": "https://api.github.com/users/sshleifer/following{/other_user}", "gists_url": "https://api.github.com/users/sshleifer/gists{/gist_id}", "starred_url": "https://api.github.com/users/sshleifer/starred{/owner}{/repo}", "subscriptions_url": "https://api.github.com/users/sshleifer/subscriptions", "organizations_url": "https://api.github.com/users/sshleifer/orgs", "repos_url": "https://api.github.com/users/sshleifer/repos", "events_url": "https://api.github.com/users/sshleifer/events{/privacy}", "received_events_url": "https://api.github.com/users/sshleifer/received_events", "type": "User", "site_admin": false } ]

[ "NVM, it's a me problem.", "FFR, You have to manually pass `use_cache=False` to bart forward now if you're not generating" ]

[ { "id": 1314768611, "node_id": "MDU6TGFiZWwxMzE0NzY4NjEx", "url": "https://api.github.com/repos/huggingface/transformers/labels/wontfix", "name": "wontfix", "color": "ffffff", "default": true, "description": null } ]

[ { "login": "patrickvonplaten", "id": 23423619, "node_id": "MDQ6VXNlcjIzNDIzNjE5", "avatar_url": "https://avatars.githubusercontent.com/u/23423619?v=4", "gravatar_id": "", "url": "https://api.github.com/users/patrickvonplaten", "html_url": "https://github.com/patrickvonplaten", "followers_url": "https://api.github.com/users/patrickvonplaten/followers", "following_url": "https://api.github.com/users/patrickvonplaten/following{/other_user}", "gists_url": "https://api.github.com/users/patrickvonplaten/gists{/gist_id}", "starred_url": "https://api.github.com/users/patrickvonplaten/starred{/owner}{/repo}", "subscriptions_url": "https://api.github.com/users/patrickvonplaten/subscriptions", "organizations_url": "https://api.github.com/users/patrickvonplaten/orgs", "repos_url": "https://api.github.com/users/patrickvonplaten/repos", "events_url": "https://api.github.com/users/patrickvonplaten/events{/privacy}", "received_events_url": "https://api.github.com/users/patrickvonplaten/received_events", "type": "User", "site_admin": false } ]

[ "Update: This seems relevant https://github.com/pytorch/pytorch/issues/36035", "I face the same error when using multi GPUs on Reformer model:\r\n```Traceback (most recent call last):\r\n File \"src/run_language_modeling.py\", line 305, in <module>\r\n main()\r\n File \"src/run_language_modeling.py\", line 269, in main\r\n trainer.train(model_path=model_path)\r\n File \"/project/6006286/qiwu/from_git/transformers/src/transformers/trainer.py\", line 499, in train\r\n tr_loss += self._training_step(model, inputs, optimizer)\r\n File \"/project/6006286/qiwu/from_git/transformers/src/transformers/trainer.py\", line 632, in _training_step\r\n outputs = model(**inputs)\r\n File \"/home/qiwu/protein-reformer-env/lib/python3.7/site-packages/torch/nn/modules/module.py\", line 550, in __call__\r\n result = self.forward(*input, **kwargs)\r\n File \"/home/qiwu/protein-reformer-env/lib/python3.7/site-packages/torch/nn/parallel/data_parallel.py\", line 155, in forward\r\n outputs = self.parallel_apply(replicas, inputs, kwargs)\r\n File \"/home/qiwu/protein-reformer-env/lib/python3.7/site-packages/torch/nn/parallel/data_parallel.py\", line 165, in parallel_apply\r\n return parallel_apply(replicas,wandb: Waiting for W&B process to finish, PID 20542\r\n inputs, kwargs, self.device_ids[:len(replicas)])\r\n File \"/home/qiwu/protein-reformer-env/lib/python3.7/site-packages/torch/nn/parallel/parallel_apply.py\", line 85, in parallel_apply\r\n output.reraise()\r\n File \"/home/qiwu/protein-reformer-env/lib/python3.7/site-packages/torch/_utils.py\", line 395, in reraise\r\n raise self.exc_type(msg)\r\nRuntimeError: Caught RuntimeError in replica 0 on device 0.\r\nOriginal Traceback (most recent call last):\r\n File \"/home/qiwu/protein-reformer-env/lib/python3.7/site-packages/torch/nn/parallel/parallel_apply.py\", line 60, in _worker\r\n output = module(*input, **kwargs)\r\n File \"/home/qiwu/protein-reformer-env/lib/python3.7/site-packages/torch/nn/modules/module.py\", line 550, in __call__\r\n result = self.forward(*input, **kwargs)\r\n File \"/project/6006286/qiwu/from_git/transformers/src/transformers/modeling_reformer.py\", line 1746, in forward\r\n return_tuple=return_tuple,\r\n File \"/home/qiwu/protein-reformer-env/lib/python3.7/site-packages/torch/nn/modules/module.py\", line 550, in __call__\r\n result = self.forward(*input, **kwargs)\r\n File \"/project/6006286/qiwu/from_git/transformers/src/transformers/modeling_reformer.py\", line 1610, in forward\r\n embedding_output = self.embeddings(input_ids=input_ids, position_ids=position_ids, inputs_embeds=inputs_embeds)\r\n File \"/home/qiwu/protein-reformer-env/lib/python3.7/site-packages/torch/nn/modules/module.py\", line 550, in __call__\r\n result = self.forward(*input, **kwargs)\r\n File \"/project/6006286/qiwu/from_git/transformers/src/transformers/modeling_reformer.py\", line 236, in forward\r\n position_embeddings = self.position_embeddings(position_ids)\r\n File \"/home/qiwu/protein-reformer-env/lib/python3.7/site-packages/torch/nn/modules/module.py\", line 550, in __call__\r\n result = self.forward(*input, **kwargs)\r\n File \"/project/6006286/qiwu/from_git/transformers/src/transformers/modeling_reformer.py\", line 143, in forward\r\n weights = torch.cat(broadcasted_weights, dim=-1)\r\nRuntimeError: There were no tensor arguments to this function (e.g., wandb: Program failed with code 1. Press ctrl-c to abort syncing.\r\nyou passed an empty list of Tensors), but no fallback function is registered for schema aten::_cat. This usually means that this function requires a non-empty list of Tensors. Available functions are [CPUTensorId, CUDATensorId, QuantizedCPUTensorId, VariableTensorId]\r\n```", "Out of curiosity, do you have the same error on PyTorch 1.4?", "> Out of curiosity, do you have the same error on PyTorch 1.4?\r\n\r\nI stopped my GC instance - now there are none available. Maybe someone else can check?", "In my case there's no error using torch-1.4.0, but got a warning:\r\n```07/16/2020 11:58:13 - INFO - transformers.trainer - ***** Running training *****\r\n07/16/2020 11:58:13 - INFO - transformers.trainer - Num examples = 5444\r\n07/16/2020 11:58:13 - INFO - transformers.trainer - Num Epochs = 12\r\n07/16/2020 11:58:13 - INFO - transformers.trainer - Instantaneous batch size per device = 32\r\n07/16/2020 11:58:13 - INFO - transformers.trainer - Total train batch size (w. parallel, distributed & accumulation) = 64\r\n07/16/2020 11:58:13 - INFO - transformers.trainer - Gradient Accumulation steps = 1\r\n07/16/2020 11:58:13 - INFO - transformers.trainer - Total optimization steps = 1000\r\nEpoch: 0%| | 0/12 [00:00<?, ?it/s\r\nhome/qiwu/torch-1.4/lib/python3.7/site-packages/torch/nn/parallel/_functions.py:61: \r\nUserWarning: Was asked to gather along dimension 0, but all input tensors were scalars; will instead unsqueeze and return a vector.\r\n```\r\n\r\nFound a relevant issue : https://github.com/huggingface/transformers/issues/852\r\nhttps://discuss.pytorch.org/t/how-to-fix-gathering-dim-0-warning-in-multi-gpu-dataparallel-setting/41733/2", "To be honest, I didn't check Reformer on multi-GPU yet - will note this issue down\r\n", "@qwu01, what version of transformers are you using, and do you also have tokenizers installed? I get a vague segmentation fault error when I attempt to run about the same training script as above using `torch==1.4.0`, `transformers==2.9.0`, and `tokenizers==0.7.0`.\r\n\r\n", "@jstremme I have these installed:\r\n\r\nPackage Version\r\n--------------- ---------\r\nargh 0.26.2\r\ncertifi 2020.6.20\r\nchardet 3.0.4\r\nclick 7.1.2\r\nconfigparser 5.0.0\r\ndocker-pycreds 0.4.0\r\nfilelock 3.0.12\r\ngitdb 4.0.5\r\nGitPython 3.1.7\r\ngql 0.2.0\r\ngraphql-core 1.1\r\nidna 2.10\r\njoblib 0.16.0\r\nnumpy 1.18.4\r\nnvidia-ml-py3 7.352.0\r\npackaging 20.4\r\npathtools 0.1.2\r\npip 19.1.1\r\npromise 2.3\r\npsutil 5.7.0\r\npyparsing 2.4.7\r\npython-dateutil 2.8.1\r\nPyYAML 5.3.1\r\nregex 2019.11.1\r\nrequests 2.24.0\r\nsacremoses 0.0.43\r\nsentencepiece 0.1.90\r\nsentry-sdk 0.16.1\r\nsetuptools 41.0.1\r\nshortuuid 1.0.1\r\nsix 1.15.0\r\nsmmap 3.0.4\r\nsubprocess32 3.5.3\r\n**tokenizers 0.8.1rc1**\r\n**torch 1.4.0**\r\ntqdm 4.47.0\r\n**transformers 3.0.2**\r\nurllib3 1.25.9\r\nwandb 0.9.3\r\nwatchdog 0.9.0\r\nwheel 0.33.4\r\n", "Thanks very much @qwu01. Just to confirm, downgrading torch to `1.4.0` allowed you to train Reformer with multiple GPUs? Did this impact anything else? \r\n\r\nThe environment I'm using does not allow me to install `tokenizers 0.8.1rc1` and `transformers 3.0.2` currently, but I will test your environment config as soon as I'm able :) \r\n\r\nIf you are actively working on training a large Reformer model, I would be interested in discussing your parameters. I'm dealing with sequences of max length around 300k SentencePiece tokens and am limited to batch size = 1. Multi-GPU should get me to batch size = 4.", "@jstremme Yes, I'm sure that torch 1.4.0 with multiple GPUs worked for Reformer training. AFAICT it's not impacting anything else.\r\n", "@qwu01, @anthonyfuller7, downgrading to `torch==1.4.0` worked for me as well :D", "This issue has been automatically marked as stale because it has not had recent activity. It will be closed if no further activity occurs. Thank you for your contributions.\n", "Was this issue ever solved? I have managed to use multiple GPUs in Reformer training by downgrading to PyTorch 1.4.0 and transformers 3.0.2. However, I would like to not be constrained to this version setup, because it is leading to some inefficiencies (functions for which arguments have changed in the new version, etc.) and also because I'd like the version to be up to date.", "@anthonyfuller7, perhaps you could reopen this? I'm in a similar position to @JellePiepenbrock where having to use `torch==1.4.0` is a suboptimal workaround.", "Sorry guys that this was never really solved -> could you try to post the problem on the forum: https://discuss.huggingface.co/ instead? It's usually more active for problems with multi-gpu training", "Sure, @patrickvonplaten. I created a [post here](https://discuss.huggingface.co/t/reformer-for-multi-gpu-not-possible-for-torch-1-4-0/9422?u=jstremme). Is there someone from Hugging Face who would be able to help resolve this? As mentioned in my post, I'd be happy to help, but I don't think I understand the code well enough to spearhead the fix. Thanks for your reply to my comment!", "> \r\n\r\nHello @jstremme \r\n\r\nI tried to downgrade to Pytorch 1.4.0 to remove the warning (and possibly increase training speed, as mentioned [here](https://github.com/huggingface/transformers/issues/852)) but I got this error:\r\n\r\n```\r\nTraceback (most recent call last):\r\n File \"run_mlm_arrow_dataset.py\", line 552, in <module>\r\n main()\r\n File \"run_mlm_arrow_dataset.py\", line 501, in main\r\n train_result = trainer.train(resume_from_checkpoint=checkpoint)\r\n File \"/home/claudio/python-environments/test_venv/lib/python3.6/site-packages/transformers/trainer.py\", line 1214, in train\r\n self.create_optimizer_and_scheduler(num_training_steps=max_steps)\r\n File \"/home/claudio/python-environments/test_venv/lib/python3.6/site-packages/transformers/trainer.py\", line 803, in create_optimizer_and_scheduler\r\n self.create_optimizer()\r\n File \"/home/claudio/python-environments/test_venv/lib/python3.6/site-packages/transformers/trainer.py\", line 836, in create_optimizer\r\n self.optimizer = optimizer_cls(optimizer_grouped_parameters, **optimizer_kwargs)\r\n File \"/home/claudio/python-environments/test_venv/lib/python3.6/site-packages/transformers/optimization.py\", line 311, in __init__\r\n require_version(\"torch>=1.5.0\") # add_ with alpha\r\n File \"/home/claudio/python-environments/test_venv/lib/python3.6/site-packages/transformers/utils/versions.py\", line 114, in require_version\r\n _compare_versions(op, got_ver, want_ver, requirement, pkg, hint)\r\n File \"/home/claudio/python-environments/test_venv/lib/python3.6/site-packages/transformers/utils/versions.py\", line 50, in _compare_versions\r\n f\"{requirement} is required for a normal functioning of this module, but found {pkg}=={got_ver}.{hint}\"\r\nImportError: torch>=1.5.0 is required for a normal functioning of this module, but found torch==1.4.0.\r\n```\r\n\r\nDid you find some solution to the issue?" ]

[]

[]

[ "I will be off for the next two weeks - maybe @sshleifer @sgugger @julien-c can take a look? ", "@JetRunner Do you know who from @microsoft uploaded the MiniLM model?", "@patrickvonplaten did it if I remember it right. He's on vocation so I'll take a look.\n", "Here's the diff\r\n\r\n@patrickvonplaten when u are back to work, pls check why this happened.\r\nI'll re-upload `vocab.txt` to resolve the problem for now.", "@julien-c I've re-uploaded it. However, CDN seems to have cached the incorrect version (https://cdn.huggingface.co/microsoft/MiniLM-L12-H384-uncased/vocab.txt).", "Yes, the CDN caches files for up to 24 hours on each POP. However AFAIK the library doesn't load tokenizer files from the CDN anyways.", "The model is working now" ]

[ { "id": 1838412367, "node_id": "MDU6TGFiZWwxODM4NDEyMzY3", "url": "https://api.github.com/repos/huggingface/transformers/labels/model%20card", "name": "model card", "color": "92d5f4", "default": false, "description": "Related to pretrained model cards" } ]

[]

[]

[]

[]

[ "There seem to be some issues with CircleCI right now causing all the integration tests to fail. Please let me know if there is an issue on my end." ]

[]

[]

[]

[]

[]

[ "Fixes for the failing tests are not detected by Github while being present on the branch ... Lets see if it comes back to life in a while ...", "CI failure has been fixed on master.", "# [Codecov](https://codecov.io/gh/huggingface/transformers/pull/5773?src=pr&el=h1) Report\n> Merging [#5773](https://codecov.io/gh/huggingface/transformers/pull/5773?src=pr&el=desc) into [master](https://codecov.io/gh/huggingface/transformers/commit/223bad242d0d64e20a39b956b73ab300231a9c70&el=desc) will **decrease** coverage by `0.04%`.\n> The diff coverage is `90.00%`.\n\n[](https://codecov.io/gh/huggingface/transformers/pull/5773?src=pr&el=tree)\n\n```diff\n@@ Coverage Diff @@\n## master #5773 +/- ##\n==========================================\n- Coverage 78.67% 78.63% -0.05% \n==========================================\n Files 146 146 \n Lines 26210 26206 -4 \n==========================================\n- Hits 20621 20606 -15 \n- Misses 5589 5600 +11 \n```\n\n\n| [Impacted Files](https://codecov.io/gh/huggingface/transformers/pull/5773?src=pr&el=tree) | Coverage Δ | |\n|---|---|---|\n| [src/transformers/modeling\\_mobilebert.py](https://codecov.io/gh/huggingface/transformers/pull/5773/diff?src=pr&el=tree#diff-c3JjL3RyYW5zZm9ybWVycy9tb2RlbGluZ19tb2JpbGViZXJ0LnB5) | `89.44% <66.66%> (-0.02%)` | :arrow_down: |\n| [src/transformers/modeling\\_bert.py](https://codecov.io/gh/huggingface/transformers/pull/5773/diff?src=pr&el=tree#diff-c3JjL3RyYW5zZm9ybWVycy9tb2RlbGluZ19iZXJ0LnB5) | `88.37% <100.00%> (-0.02%)` | :arrow_down: |\n| [src/transformers/modeling\\_openai.py](https://codecov.io/gh/huggingface/transformers/pull/5773/diff?src=pr&el=tree#diff-c3JjL3RyYW5zZm9ybWVycy9tb2RlbGluZ19vcGVuYWkucHk=) | `82.25% <100.00%> (-0.06%)` | :arrow_down: |\n| [src/transformers/modeling\\_xlm.py](https://codecov.io/gh/huggingface/transformers/pull/5773/diff?src=pr&el=tree#diff-c3JjL3RyYW5zZm9ybWVycy9tb2RlbGluZ194bG0ucHk=) | `90.00% <100.00%> (-0.03%)` | :arrow_down: |\n| [src/transformers/generation\\_tf\\_utils.py](https://codecov.io/gh/huggingface/transformers/pull/5773/diff?src=pr&el=tree#diff-c3JjL3RyYW5zZm9ybWVycy9nZW5lcmF0aW9uX3RmX3V0aWxzLnB5) | `83.95% <0.00%> (-2.51%)` | :arrow_down: |\n| [src/transformers/file\\_utils.py](https://codecov.io/gh/huggingface/transformers/pull/5773/diff?src=pr&el=tree#diff-c3JjL3RyYW5zZm9ybWVycy9maWxlX3V0aWxzLnB5) | `81.19% <0.00%> (-0.30%)` | :arrow_down: |\n\n------\n\n[Continue to review full report at Codecov](https://codecov.io/gh/huggingface/transformers/pull/5773?src=pr&el=continue).\n> **Legend** - [Click here to learn more](https://docs.codecov.io/docs/codecov-delta)\n> `Δ = absolute <relative> (impact)`, `ø = not affected`, `? = missing data`\n> Powered by [Codecov](https://codecov.io/gh/huggingface/transformers/pull/5773?src=pr&el=footer). Last update [b01a884...722824f](https://codecov.io/gh/huggingface/transformers/pull/5773?src=pr&el=lastupdated). Read the [comment docs](https://docs.codecov.io/docs/pull-request-comments).\n", "@sshleifer I've updated all the models with `position_ids` created in the forward pass as it seems to be the right design to allow people to export the bunch of models impacted. \r\n\r\nAlso, it avoids allocating tensors at every forward call, so might reduce the pressure on PyTorch memory manager. ", "Do we get the \"Uninitialized parameters\" warning because the position_ids are not in the `pytorch_model.bin` on S3?\r\n\r\n", "I'll check the above, didn't catch at first 👌 ", "Could this be merged? Would love to test these changes after merging :)", "@vdantu Now merged, sorry for the delay, many people are off these days, it slows down a little bit the merging process 😃. \r\n\r\nLet us know if you have any follow up issue(s) / question(s).", "@mfuntowicz : Thanks for the push. Would we have to wait for a new release of `transformers` package to get these changes? ", "@vdantu: I'll let @LysandreJik handle this one " ]

[]

[]

[ "Merging without circleci." ]

[ { "id": 1314768611, "node_id": "MDU6TGFiZWwxMzE0NzY4NjEx", "url": "https://api.github.com/repos/huggingface/transformers/labels/wontfix", "name": "wontfix", "color": "ffffff", "default": true, "description": null } ]

[]

[ "Hey @rxlian, \r\n\r\nWe are trying to move more general and \"researchy\" questions to our discussion forum here: https://discuss.huggingface.co/ and use github issues mainly for bugs. \r\n\r\nWould you mind posting your questions at the forum - other people might be interested as well :-) ", "This issue has been automatically marked as stale because it has not had recent activity. It will be closed if no further activity occurs. Thank you for your contributions.\n" ]

[]

[]

[ "# [Codecov](https://codecov.io/gh/huggingface/transformers/pull/5770?src=pr&el=h1) Report\n> Merging [#5770](https://codecov.io/gh/huggingface/transformers/pull/5770?src=pr&el=desc) into [master](https://codecov.io/gh/huggingface/transformers/commit/3653d01f2af0389207f2239875a8ceae41bf0598&el=desc) will **decrease** coverage by `0.01%`.\n> The diff coverage is `100.00%`.\n\n[](https://codecov.io/gh/huggingface/transformers/pull/5770?src=pr&el=tree)\n\n```diff\n@@ Coverage Diff @@\n## master #5770 +/- ##\n==========================================\n- Coverage 77.26% 77.25% -0.02% \n==========================================\n Files 146 146 \n Lines 25948 25958 +10 \n==========================================\n+ Hits 20048 20053 +5 \n- Misses 5900 5905 +5 \n```\n\n\n| [Impacted Files](https://codecov.io/gh/huggingface/transformers/pull/5770?src=pr&el=tree) | Coverage Δ | |\n|---|---|---|\n| [src/transformers/configuration\\_utils.py](https://codecov.io/gh/huggingface/transformers/pull/5770/diff?src=pr&el=tree#diff-c3JjL3RyYW5zZm9ybWVycy9jb25maWd1cmF0aW9uX3V0aWxzLnB5) | `96.52% <ø> (ø)` | |\n| [src/transformers/configuration\\_xlnet.py](https://codecov.io/gh/huggingface/transformers/pull/5770/diff?src=pr&el=tree#diff-c3JjL3RyYW5zZm9ybWVycy9jb25maWd1cmF0aW9uX3hsbmV0LnB5) | `94.33% <ø> (ø)` | |\n| [src/transformers/modeling\\_xlnet.py](https://codecov.io/gh/huggingface/transformers/pull/5770/diff?src=pr&el=tree#diff-c3JjL3RyYW5zZm9ybWVycy9tb2RlbGluZ194bG5ldC5weQ==) | `81.75% <100.00%> (+0.26%)` | :arrow_up: |\n| [src/transformers/generation\\_tf\\_utils.py](https://codecov.io/gh/huggingface/transformers/pull/5770/diff?src=pr&el=tree#diff-c3JjL3RyYW5zZm9ybWVycy9nZW5lcmF0aW9uX3RmX3V0aWxzLnB5) | `85.21% <0.00%> (-1.26%)` | :arrow_down: |\n\n------\n\n[Continue to review full report at Codecov](https://codecov.io/gh/huggingface/transformers/pull/5770?src=pr&el=continue).\n> **Legend** - [Click here to learn more](https://docs.codecov.io/docs/codecov-delta)\n> `Δ = absolute <relative> (impact)`, `ø = not affected`, `? = missing data`\n> Powered by [Codecov](https://codecov.io/gh/huggingface/transformers/pull/5770?src=pr&el=footer). Last update [3653d01...cfd7c66](https://codecov.io/gh/huggingface/transformers/pull/5770?src=pr&el=lastupdated). Read the [comment docs](https://docs.codecov.io/docs/pull-request-comments).\n", "Not sure what is going on with Cicle CI here - might be related to the problems with github yesterday...", "Yeah it doesn't want to let me re-run, I think I'll just add a small change to refresh it", "After our discussion yesterday @TevenLeScao, I think I am more or less fine with this PR. A couple of things we should maybe think about:\r\n\r\n1) ***Consistency with `use_cache` of other models***\r\nAs it is implemented at the moment (correct me if I'm wrong), if `use_cache=False`, `mem_len` has no effect as `mems` will always be `None`. This is not ideal IMO because this means that one has to enable `use_cache=True` to correctly train the model => this is not really consistent with the way `use_cache` is used in the library in general. In all other models GPT2, CTRL, T5, BART, Reformer `use_cache` should only be used for inference and disabled for training. So as a solution I think, there are two options:\r\n A.) All `use_cache` statements should be updated with `use_cache or self.training` => this way I think it would be consistent with the logic / design of other models (Let me know if this is not clear)\r\n B.) Add an explicit warning that `use_cache` should be `True` for training to enable the `mems` => I don't like this very much because this is confusing for people that are used to GPT2 where `use_cache` is only for inference. \r\n\r\n2) ***Equality `use_cache=True/False`*** \r\nAfter our discussion yesterday, it seems that since `XLNet` uses a mixture of CLM and bi-directional self-attention (defined in the `perm_mask` input), it seems that it is **not** possible to have a (mathematically) identical output between `use_cache=True` and `use_cache=False`. I guess that's just how it is in `XLNet` and we can't really do anything against it. \r\n\r\n - A.) I guess we should change (...could be done in a 2nd PR though), is to add the parameter `offset` to `XLNet` config and add it to all `forward(...)` functions and maybe give it a better name... As I understood it yesterday, `offset` defines the trade-off between how many tokens should be used from `mems` (and are therefore not in the *query projection*) and how many are in the query projection and have a bi-directional mask on them. At the moment `offset` is hard-coded to `2` in the `prepare_generate` function which does not allow for a lot of flexibility and is somewhat arbitrary IMO. I think, this parameter should be handled in a similar way as `num_hashes` is handled in `Reformer`\r\n\r\n- B.) This is one is more important to me. We should add a test to `XLNet` that verifies that `use_cache` does give the **same** outputs if the `perm_mask` is set to a causal mask. We have tests for this in `GPT2, T5, Bart` I think. Here is the one for GPT2 e.g.: https://github.com/huggingface/transformers/blob/1f75f9317e381425ee56f7108e5ec8d3f3d6b6ad/tests/test_modeling_gpt2.py#L165 . The test here can be very similar - only that we need to define the `perm_mask` correctly. This test would be extremely useful a) to make sure we fully understand what's going on in `XLNet` and b) make sure that there is no bug.\r\n\r\n3) ***Benchmark performance improvement***\r\n\r\nSoon we will have benchmarking for the `generate` function, see PR here: https://github.com/huggingface/transformers/pull/5802 . So I took this case here as a good trial to test the benchmark scripts for generation. I started two benchmarks 1 on CPU 1 on GPU and will post the results here as soon as the benchmarking is done. It be great to add benchmark results in general to PRs like this.\r\n\r\n\r\nI think this is quite an important Model and an important PR, so I'd love to loop in @sshleifer and @thomwolf here as well to hear their opinions. \r\n\r\n@TevenLeScao - sorry for the long message! It's mostly because I want to better understand `XLNet` in detail. Let me know if some of my points are not clear or not 100% correct.", "I used the following code for benchmarking:\r\n\r\n```python\r\n\r\n#!/usr/bin/env python3\r\nfrom transformers import XLNetConfig, PyTorchBenchmark, PyTorchBenchmarkArguments\r\n\r\nconfig_with_cache = XLNetConfig.from_pretrained(\"xlnet-base-cased\", use_cache=True, mem_len=1024)\r\nconfig_without_cache = XLNetConfig.from_pretrained(\"xlnet-base-cased\", use_cache=False, mem_len=1024)\r\nconfig_without_cache_no_mems = XLNetConfig.from_pretrained(\"xlnet-base-cased\", use_cache=False)\r\n\r\nassert config_without_cache_no_mems.mem_len is None, \"Configs are wrong\"\r\nassert config_with_cache.mem_len == config_without_cache.mem_len == 1024, \"Configs are wrong\"\r\nassert config_with_cache.use_cache is (not config_without_cache.use_cache) is True and (not config_without_cache_no_mems.use_cache) is True, \"Configs are wrong\"\r\n\r\nargs = PyTorchBenchmarkArguments(models=[\"xlnet-cache\", \"xlnet-no-cache\", \"xlnet-no-mems\"], sequence_lengths=[300], batch_sizes=[1], generate=True, no_inference=True)\r\n\r\nbenchmark = PyTorchBenchmark(args=args, configs=[config_with_cache, config_without_cache, config_without_cache_no_mems])\r\nbenchmark.run()\r\n\r\n\r\n```\r\nand the benchmark code from #5802 \r\n\r\n\r\n=> I could not see a real speed-up on GPU for the following case:\r\nStart with 1 token and generate up to `Seq Length` tokens:\r\n\r\n INFERENCE - SPEED - RESULT\r\n--------------------------------------------------------------------------------\r\n Model Name Batch Size Seq Length Time in s \r\n--------------------------------------------------------------------------------\r\n xlnet-cache 1 500 11.8 \r\n xlnet-no-cache 1 500 11.925 \r\n xlnet-no-mems 1 500 11.911 \r\n--------------------------------------------------------------------------------\r\n\r\n INFERENCE - MEMORY - RESULT \r\n--------------------------------------------------------------------------------\r\n Model Name Batch Size Seq Length Memory in MB \r\n--------------------------------------------------------------------------------\r\n xlnet-cache 1 500 1351 \r\n xlnet-no-cache 1 500 1397 \r\n xlnet-no-mems 1 500 1397 \r\n--------------------------------------------------------------------------------\r\n\r\n ENVIRONMENT INFORMATION \r\n- transformers_version: 3.0.2\r\n- framework: PyTorch\r\n- use_torchscript: False\r\n- framework_version: 1.5.0\r\n- python_version: 3.7.7\r\n- system: Linux\r\n- cpu: x86_64\r\n- architecture: 64bit\r\n- date: 2020-07-16\r\n- time: 11:28:43.644495\r\n- fp16: False\r\n- use_multiprocessing: True\r\n- only_pretrain_model: False\r\n- cpu_ram_mb: 32089\r\n- use_gpu: True\r\n- num_gpus: 1\r\n- gpu: TITAN RTX\r\n- gpu_ram_mb: 24217\r\n- gpu_power_watts: 280.0\r\n- gpu_performance_state: 0\r\n- use_tpu: False\r\n\r\n=> For CPU, there is a small speed up and the speed up is probably more significant for longer sequence length (so I run it again for longer sequences):\r\n\r\nINFERENCE - SPEED - RESULT\r\n--------------------------------------------------------------------------------\r\n Model Name Batch Size Seq Length Time in s \r\n--------------------------------------------------------------------------------\r\n xlnet-cache 1 50 5.032 \r\n xlnet-no-cache 1 50 6.149 \r\n xlnet-no-mems 1 50 6.175 \r\n--------------------------------------------------------------------------------\r\n\r\nINFERENCE - MEMORY - RESULT \r\n--------------------------------------------------------------------------------\r\n Model Name Batch Size Seq Length Memory in MB \r\n--------------------------------------------------------------------------------\r\n xlnet-cache 1 50 691 \r\n xlnet-no-cache 1 50 695 \r\n xlnet-no-mems 1 50 696 \r\n--------------------------------------------------------------------------------\r\n\r\nENVIRONMENT INFORMATION\r\n- transformers_version: 3.0.2\r\n- framework: PyTorch\r\n- use_torchscript: False\r\n- framework_version: 1.5.0\r\n- python_version: 3.7.7\r\n- system: Linux\r\n- cpu: x86_64\r\n- architecture: 64bit\r\n- date: 2020-07-16\r\n- time: 11:28:43.644495\r\n- fp16: False\r\n- use_multiprocessing: True\r\n- only_pretrain_model: False\r\n- cpu_ram_mb: 32089\r\n- use_gpu: False\r\n- use_tpu: False\r\n\r\n\r\nBut overall, the speed-up is not really significant... @TevenLeScao can you post the code you used to benchmark the CPU speed-up here? Maybe I am doing something wrong.\r\n\r\nI think if we manage to cache the key, value projections instead of recalculating them for every token we would see a bigger speed-up, but this has to checked.", "- on GPU I also haven't observed a difference, and doing inference/text generation on 1 sentence isn't going to be very affected by the caching trick since the GPU is going to parallelize everything anyway\r\n- on CPU I've run from the text generation pipeline:\r\n\r\n```\r\n#!/usr/bin/env python3\r\nfrom transformers import pipeline, XLNetLMHeadModel\r\nimport time\r\n\r\n# xlnet = XLNetLMHeadModel.from_pretrained(\"xlnet-base-cased\")\r\nxlnet = XLNetLMHeadModel.from_pretrained(\"xlnet-base-cased\", use_cache=False)\r\ngenerator = pipeline(\"text-generation\", model=xlnet, tokenizer=\"xlnet-base-cased\")\r\nstart = time.time()\r\nnumber_gen = 100\r\nfor i in range(number_gen):\r\n output_to_check = generator(\"Today is a beautiful day and I, \")\r\n print(output_to_check)\r\n print((time.time() - start) / (i+1))\r\n print()\r\n```\r\n\r\nwhich makes I think a big (33.2 vs 8.6 seconds on my CPU) difference since XLNet has the 170-long padding text at the start! Which we should probably also take a look at since it heavily influences the generated text...", "I can confirm a much bigger speed-up on CPU for longer sequences. Here the results:\r\n\r\n==================== INFERENCE - SPEED - RESULT ==================== \r\n-------------------------------------------------------------------------------- \r\n Model Name Batch Size Seq Length Time in s \r\n--------------------------------------------------------------------------------\r\n xlnet-cache 1 300 37.733 \r\n xlnet-no-cache 1 300 98.142 \r\n xlnet-no-mems 1 300 94.338 \r\n-------------------------------------------------------------------------------- \r\n \r\n==================== INFERENCE - MEMORY - RESULT ====================\r\n--------------------------------------------------------------------------------\r\n Model Name Batch Size Seq Length Memory in MB \r\n-------------------------------------------------------------------------------- \r\n xlnet-cache 1 300 728 \r\n xlnet-no-cache 1 300 752 \r\n xlnet-no-mems 1 300 750 \r\n-------------------------------------------------------------------------------- \r\n \r\n==================== ENVIRONMENT INFORMATION ==================== \r\n- transformers_version: 3.0.2 \r\n- framework: PyTorch \r\n- use_torchscript: False \r\n- framework_version: 1.5.0 \r\n- python_version: 3.7.7\r\n- system: Linux\r\n- cpu: x86_64\r\n- architecture: 64bit \r\n- date: 2020-07-16 \r\n- time: 13:49:43.409581 \r\n- fp16: False \r\n- use_multiprocessing: True \r\n- only_pretrain_model: False \r\n- cpu_ram_mb: 32089 \r\n- use_gpu: False \r\n- use_tpu: False\r\n", "@patrickvonplaten I added the `self.training` check and the test for `use_cache` in autoregressive mode.", "So weird, it passed the tests earlier but the CI just told me the push had been refused, I'll revert it and look at it again tomorrow" ]

[]

[]

[ "Code Quality is failing due to files that were not modified in this PR, not addressing to avoid rebasing somewhere else.", "ill fix code quality." ]

[ { "id": 1314768611, "node_id": "MDU6TGFiZWwxMzE0NzY4NjEx", "url": "https://api.github.com/repos/huggingface/transformers/labels/wontfix", "name": "wontfix", "color": "ffffff", "default": true, "description": null } ]

[]

[ "This issue has been automatically marked as stale because it has not had recent activity. It will be closed if no further activity occurs. Thank you for your contributions.\n", "Have you got your answer? I got the same question.." ]

[ { "id": 1314768611, "node_id": "MDU6TGFiZWwxMzE0NzY4NjEx", "url": "https://api.github.com/repos/huggingface/transformers/labels/wontfix", "name": "wontfix", "color": "ffffff", "default": true, "description": null } ]

[]

[ "It seems you're trying to identify entities in your query? You could use an NER model to do so, but I'm not sure the entities the NER models on the hub are trained on would work in your use-case. You can still check them out, and check the [NER script](https://github.com/huggingface/transformers/tree/master/examples/token-classification) if you want to train a model with your data.", "This issue has been automatically marked as stale because it has not had recent activity. It will be closed if no further activity occurs. Thank you for your contributions.\n" ]

[]

[]

[ "Hi @BCWang93, I had the same issue, you need to have scikit-learn installed to run 'run_glue.py'.\r\n\r\n`pip install scikit-learn` solved the issue for me.", "Indeed, this issue happens when `scikit-learn` is not installed. Thanks @nassim-yagoub!" ]

[]

[]

[ "Hello!\r\n\r\nCan you give a code that better explain your issue and a way to reproduce it please? Thanks :) ", "Sorry! I check again and it works well!\r\nClose it!\r\n" ]

[]

[]

[ "In `RUN_SLOW=1` the new tests: `test_saved_model_with_attentions_outputs` and `test_saved_model_with_hidden_states_output` fail @jplu from https://github.com/huggingface/transformers/pull/5468. The problem is that I have to use `tf.cond(...)` and it seems like this forces me to also use `cast_bool_...`. Not sure if you have any ideas on how to fix this @jplu .", "Yes, It is still an issue with the AutoGraph thing :cry: I suggest to comment them for now.", "I'm currently thinking on how to properly rework all the booleans handling in TF. As this is the main issue.", "Ok...will leave it for now since the test are only in `RUN_SLOW` mode, so they won't show up in Circle CI", "Confirmed that this PR will not slow down the Longformer PyTorch version.\r\n\r\nRunning the following benchmark on master:\r\n```\r\npython examples/benchmarking/run_benchmark.py --models allenai/longformer-base-4096 --no_memory --sequence_length 512 1024\r\n```\r\ngives same performance is in as in https://github.com/huggingface/transformers/pull/5811.", "Benchmarking the model in TF leads to a slow-down vs. PyTorch of ca. 1.5, which is reasonable IMO:\r\n\r\nRunning: \r\n```\r\npython examples/benchmarking/run_benchmark_tf.py --models allenai/longformer-base-4096 --no_memory --sequence_length 512 1024\r\n```\r\n\r\ngives:\r\n\r\n```\r\n \r\n==================== INFERENCE - SPEED - RESULT ==================== \r\n-------------------------------------------------------------------------------- \r\n Model Name Batch Size Seq Length Time in s \r\n-------------------------------------------------------------------------------- \r\n allenai/longformer-base-4096 8 512 0.226 \r\n allenai/longformer-base-4096 8 1024 0.446 \r\n-------------------------------------------------------------------------------- \r\n\r\n==================== ENVIRONMENT INFORMATION ====================\r\n- transformers_version: 3.0.2\r\n- framework: TensorFlow\r\n- eager_mode: False\r\n- use_xla: False\r\n- framework_version: 2.3.0\r\n- python_version: 3.6.10\r\n- system: Linux\r\n- cpu: x86_64\r\n- architecture: 64bit\r\n- date: 2020-08-06\r\n- time: 15:55:55.754100\r\n- fp16: False\r\n- use_multiprocessing: True\r\n- only_pretrain_model: False\r\n- cpu_ram_mb: 32088\r\n- use_gpu: True\r\n- num_gpus: 1\r\n- gpu: TITAN RTX\r\n- gpu_ram_mb: 24217\r\n- gpu_power_watts: 280.0\r\n- gpu_performance_state: 0\r\n- use_tpu: False\r\n```\r\n\r\nAt the moment running the model in XLA on GPU fails...=> should take a closer look in a next PR.", "# [Codecov](https://codecov.io/gh/huggingface/transformers/pull/5764?src=pr&el=h1) Report\n> Merging [#5764](https://codecov.io/gh/huggingface/transformers/pull/5764?src=pr&el=desc) into [master](https://codecov.io/gh/huggingface/transformers/commit/1b8a7ffcfdfe37f5440ac0eafb58089ff5aef00a&el=desc) will **increase** coverage by `0.05%`.\n> The diff coverage is `24.52%`.\n\n[](https://codecov.io/gh/huggingface/transformers/pull/5764?src=pr&el=tree)\n\n```diff\n@@ Coverage Diff @@\n## master #5764 +/- ##\n==========================================\n+ Coverage 79.33% 79.38% +0.05% \n==========================================\n Files 148 149 +1 \n Lines 27196 27670 +474 \n==========================================\n+ Hits 21577 21967 +390 \n- Misses 5619 5703 +84 \n```\n\n\n| [Impacted Files](https://codecov.io/gh/huggingface/transformers/pull/5764?src=pr&el=tree) | Coverage Δ | |\n|---|---|---|\n| [src/transformers/modeling\\_tf\\_longformer.py](https://codecov.io/gh/huggingface/transformers/pull/5764/diff?src=pr&el=tree#diff-c3JjL3RyYW5zZm9ybWVycy9tb2RlbGluZ190Zl9sb25nZm9ybWVyLnB5) | `15.76% <15.76%> (ø)` | |\n| [src/transformers/modeling\\_longformer.py](https://codecov.io/gh/huggingface/transformers/pull/5764/diff?src=pr&el=tree#diff-c3JjL3RyYW5zZm9ybWVycy9tb2RlbGluZ19sb25nZm9ybWVyLnB5) | `89.76% <92.98%> (+0.54%)` | :arrow_up: |\n| [src/transformers/\\_\\_init\\_\\_.py](https://codecov.io/gh/huggingface/transformers/pull/5764/diff?src=pr&el=tree#diff-c3JjL3RyYW5zZm9ybWVycy9fX2luaXRfXy5weQ==) | `99.25% <100.00%> (+<0.01%)` | :arrow_up: |\n| [src/transformers/modeling\\_tf\\_auto.py](https://codecov.io/gh/huggingface/transformers/pull/5764/diff?src=pr&el=tree#diff-c3JjL3RyYW5zZm9ybWVycy9tb2RlbGluZ190Zl9hdXRvLnB5) | `66.86% <100.00%> (+0.20%)` | :arrow_up: |\n| [src/transformers/modeling\\_tf\\_flaubert.py](https://codecov.io/gh/huggingface/transformers/pull/5764/diff?src=pr&el=tree#diff-c3JjL3RyYW5zZm9ybWVycy9tb2RlbGluZ190Zl9mbGF1YmVydC5weQ==) | `24.53% <0.00%> (-63.20%)` | :arrow_down: |\n| [src/transformers/tokenization\\_bert.py](https://codecov.io/gh/huggingface/transformers/pull/5764/diff?src=pr&el=tree#diff-c3JjL3RyYW5zZm9ybWVycy90b2tlbml6YXRpb25fYmVydC5weQ==) | `92.41% <0.00%> (+1.33%)` | :arrow_up: |\n| [src/transformers/tokenization\\_openai.py](https://codecov.io/gh/huggingface/transformers/pull/5764/diff?src=pr&el=tree#diff-c3JjL3RyYW5zZm9ybWVycy90b2tlbml6YXRpb25fb3BlbmFpLnB5) | `84.09% <0.00%> (+1.51%)` | :arrow_up: |\n| [src/transformers/tokenization\\_utils.py](https://codecov.io/gh/huggingface/transformers/pull/5764/diff?src=pr&el=tree#diff-c3JjL3RyYW5zZm9ybWVycy90b2tlbml6YXRpb25fdXRpbHMucHk=) | `91.60% <0.00%> (+1.59%)` | :arrow_up: |\n| ... and [7 more](https://codecov.io/gh/huggingface/transformers/pull/5764/diff?src=pr&el=tree-more) | |\n\n------\n\n[Continue to review full report at Codecov](https://codecov.io/gh/huggingface/transformers/pull/5764?src=pr&el=continue).\n> **Legend** - [Click here to learn more](https://docs.codecov.io/docs/codecov-delta)\n> `Δ = absolute <relative> (impact)`, `ø = not affected`, `? = missing data`\n> Powered by [Codecov](https://codecov.io/gh/huggingface/transformers/pull/5764?src=pr&el=footer). Last update [1b8a7ff...41cb64f](https://codecov.io/gh/huggingface/transformers/pull/5764?src=pr&el=lastupdated). Read the [comment docs](https://docs.codecov.io/docs/pull-request-comments).\n", "Where do I send the pizza & beer to get this merged?", "I'm curious if you think the longformer should be added to the language_generation_model? ", "@sgugger - I added the `return_dict` functionality and adapted the doc strings -> the new docstring functions are awesome!", "> I'm curious if you think the longformer should be added to the language_generation_model?\r\n\r\nWe will add it to the EncoderDecoderModel framework, where it can be used with `generate()`" ]

[ { "id": 1838412367, "node_id": "MDU6TGFiZWwxODM4NDEyMzY3", "url": "https://api.github.com/repos/huggingface/transformers/labels/model%20card", "name": "model card", "color": "92d5f4", "default": false, "description": "Related to pretrained model cards" } ]

[]

[ "We shouldn't have to add those because the library should work out of the box with the models at https://huggingface.co/nghuyong\r\n\r\nThe discoverability of those models on huggingface.co is a different thing and we're open to suggestions to improve it\r\n\r\nFor instance, you should upload model cards describing the models including metadata etc. See https://huggingface.co/docs for instance\r\n\r\nPinging @JetRunner and @sshleifer for feedback", "Thanks for your contribution!\n\nYep, I think this PR is not very necessary here since ERNIE serves as weights for BERT. We have shifted into a \"model hub\" model so I can't think of a reason to keep hard-coding urls here (and we think we should even remove the existing ones at some point) @julien-c ", "Also, what do you mean by \"especially on Chinese\"? Aren't these checkpoints in English I think?", "We should add some docs and tweet!", "You could copy https://github.com/nghuyong/ERNIE-Pytorch/blob/master/Readme.md to each model card.\r\n", "@JetRunner \r\nERNIE1.0 is for Chinese, more performance detail: https://arxiv.org/abs/1904.09223\r\nERNIE2.0 and ERNIE-tiny are for English.", "> @JetRunner \n> \n> ERNIE1.0 is for Chinese, more performance detail: https://arxiv.org/abs/1904.09223\n> \n> ERNIE2.0 and ERNIE-tiny are for English.\n\nYeah, this raises some confusion here. Thanks for your clarification!", "@julien-c @JetRunner @sshleifer \r\nI have withdrew my previous submission, and add model_card in this new commit.", "# [Codecov](https://codecov.io/gh/huggingface/transformers/pull/5763?src=pr&el=h1) Report\n> Merging [#5763](https://codecov.io/gh/huggingface/transformers/pull/5763?src=pr&el=desc) into [master](https://codecov.io/gh/huggingface/transformers/commit/8ab565a4be5a7fd96b19ef88d474037ef31f27e5&el=desc) will **decrease** coverage by `0.08%`.\n> The diff coverage is `n/a`.\n\n[](https://codecov.io/gh/huggingface/transformers/pull/5763?src=pr&el=tree)\n\n```diff\n@@ Coverage Diff @@\n## master #5763 +/- ##\n==========================================\n- Coverage 77.32% 77.24% -0.09% \n==========================================\n Files 146 146 \n Lines 26047 26047 \n==========================================\n- Hits 20141 20120 -21 \n- Misses 5906 5927 +21 \n```\n\n\n| [Impacted Files](https://codecov.io/gh/huggingface/transformers/pull/5763?src=pr&el=tree) | Coverage Δ | |\n|---|---|---|\n| [src/transformers/modeling\\_tf\\_distilbert.py](https://codecov.io/gh/huggingface/transformers/pull/5763/diff?src=pr&el=tree#diff-c3JjL3RyYW5zZm9ybWVycy9tb2RlbGluZ190Zl9kaXN0aWxiZXJ0LnB5) | `64.90% <0.00%> (-33.90%)` | :arrow_down: |\n| [src/transformers/generation\\_utils.py](https://codecov.io/gh/huggingface/transformers/pull/5763/diff?src=pr&el=tree#diff-c3JjL3RyYW5zZm9ybWVycy9nZW5lcmF0aW9uX3V0aWxzLnB5) | `97.11% <0.00%> (+0.28%)` | :arrow_up: |\n| [src/transformers/modeling\\_tf\\_roberta.py](https://codecov.io/gh/huggingface/transformers/pull/5763/diff?src=pr&el=tree#diff-c3JjL3RyYW5zZm9ybWVycy9tb2RlbGluZ190Zl9yb2JlcnRhLnB5) | `93.36% <0.00%> (+49.37%)` | :arrow_up: |\n\n------\n\n[Continue to review full report at Codecov](https://codecov.io/gh/huggingface/transformers/pull/5763?src=pr&el=continue).\n> **Legend** - [Click here to learn more](https://docs.codecov.io/docs/codecov-delta)\n> `Δ = absolute <relative> (impact)`, `ø = not affected`, `? = missing data`\n> Powered by [Codecov](https://codecov.io/gh/huggingface/transformers/pull/5763?src=pr&el=footer). Last update [8ab565a...0c8a23f](https://codecov.io/gh/huggingface/transformers/pull/5763?src=pr&el=lastupdated). Read the [comment docs](https://docs.codecov.io/docs/pull-request-comments).\n", "Thanks @nghuyong! The model cards look awesome.\nWe'll take it from here and add some metadata and then merge the model cards. ", "@JetRunner Thanks", "Hi @nghuyong ,\r\nI just tried [your script](https://github.com/nghuyong/ERNIE-Pytorch#reproduce-ernie-papers-case), and found that\r\n```\r\nSome weights of BertForMaskedLM were not initialized from the model checkpoint at nghuyong/ernie-1.0 and are newly initialized: ['cls.predictions.transform.LayerNorm.bias', 'cls.predictions.decoder.bias', 'cls.predictions.bias', 'cls.predictions.transform.LayerNorm.weight', 'cls.predictions.transform.dense.bias', 'cls.predictions.decoder.weight', 'cls.predictions.transform.dense.weight']\r\nYou should probably TRAIN this model on a down-stream task to be able to use it for predictions and inference.\r\n```\r\n\r\nThis error is very common if you upload a model that was loaded by `BertModel`. Could you re-upload these checkpoints by loading them in `BertForMaskedLM` first? That would be something like\r\n\r\n```python\r\ntokenizer = BertTokenizer.from_pretrained('./convert')\r\nmodel = BertForMaskedLM.from_pretrained('./convert')\r\n# instead of `model = BertModel.from_pretrained('./convert')`\r\n\r\nmodel.save_pretrained('./saved')\r\n```\r\nThis will allow users to directly use the checkpoints to do mask filling without fine-tuning them. Also, could you help convert the checkpoints to TensorFlow as well? That would be super easy. We have a tutorial here: https://huggingface.co/transformers/model_sharing.html\r\n\r\nThank you!", "@JetRunner OK，I will update soon", "@JetRunner have updated now~" ]

[ { "id": 1314768611, "node_id": "MDU6TGFiZWwxMzE0NzY4NjEx", "url": "https://api.github.com/repos/huggingface/transformers/labels/wontfix", "name": "wontfix", "color": "ffffff", "default": true, "description": null } ]

[]

[ "Is there any update on this? Thanks.", "@lucidrains made this [repo](https://github.com/lucidrains/mixture-of-experts) with me.", "Thanks!", "This issue has been automatically marked as stale because it has not had recent activity. It will be closed if no further activity occurs. Thank you for your contributions.\n" ]

[ { "id": 2139563322, "node_id": "MDU6TGFiZWwyMTM5NTYzMzIy", "url": "https://api.github.com/repos/huggingface/transformers/labels/cleanup", "name": "cleanup", "color": "e7fc49", "default": false, "description": "" } ]

[]

[]

[]

[]

[ "> The UI for inputting a list of potential labels + possibly a hypothesis template is non trivial\r\n\r\nI think for purposes of the early inference API, customizing the hypothesis template is not all that important. I added support for providing labels as a list of comma-delimited strings instead of a list.\r\n\r\n> one model is currently linked to one (and one only) Pipeline type in the inference API.\r\n\r\nWould it be possible to have one model, e.g. `bart-large-mnli`, linked to the zero shot pipeline and have the others remain linked to text classification?", "# [Codecov](https://codecov.io/gh/huggingface/transformers/pull/5760?src=pr&el=h1) Report\n> Merging [#5760](https://codecov.io/gh/huggingface/transformers/pull/5760?src=pr&el=desc) into [master](https://codecov.io/gh/huggingface/transformers/commit/8ab565a4be5a7fd96b19ef88d474037ef31f27e5&el=desc) will **increase** coverage by `0.72%`.\n> The diff coverage is `100.00%`.\n\n[](https://codecov.io/gh/huggingface/transformers/pull/5760?src=pr&el=tree)\n\n```diff\n@@ Coverage Diff @@\n## master #5760 +/- ##\n==========================================\n+ Coverage 77.32% 78.05% +0.72% \n==========================================\n Files 146 146 \n Lines 26047 26089 +42 \n==========================================\n+ Hits 20141 20363 +222 \n+ Misses 5906 5726 -180 \n```\n\n\n| [Impacted Files](https://codecov.io/gh/huggingface/transformers/pull/5760?src=pr&el=tree) | Coverage Δ | |\n|---|---|---|\n| [src/transformers/pipelines.py](https://codecov.io/gh/huggingface/transformers/pull/5760/diff?src=pr&el=tree#diff-c3JjL3RyYW5zZm9ybWVycy9waXBlbGluZXMucHk=) | `78.47% <100.00%> (+1.50%)` | :arrow_up: |\n| [src/transformers/modeling\\_tf\\_openai.py](https://codecov.io/gh/huggingface/transformers/pull/5760/diff?src=pr&el=tree#diff-c3JjL3RyYW5zZm9ybWVycy9tb2RlbGluZ190Zl9vcGVuYWkucHk=) | `20.27% <0.00%> (-74.92%)` | :arrow_down: |\n| [src/transformers/modeling\\_tf\\_distilbert.py](https://codecov.io/gh/huggingface/transformers/pull/5760/diff?src=pr&el=tree#diff-c3JjL3RyYW5zZm9ybWVycy9tb2RlbGluZ190Zl9kaXN0aWxiZXJ0LnB5) | `64.90% <0.00%> (-33.90%)` | :arrow_down: |\n| [src/transformers/modeling\\_openai.py](https://codecov.io/gh/huggingface/transformers/pull/5760/diff?src=pr&el=tree#diff-c3JjL3RyYW5zZm9ybWVycy9tb2RlbGluZ19vcGVuYWkucHk=) | `81.02% <0.00%> (-1.29%)` | :arrow_down: |\n| [src/transformers/tokenization\\_utils\\_base.py](https://codecov.io/gh/huggingface/transformers/pull/5760/diff?src=pr&el=tree#diff-c3JjL3RyYW5zZm9ybWVycy90b2tlbml6YXRpb25fdXRpbHNfYmFzZS5weQ==) | `92.44% <0.00%> (-1.17%)` | :arrow_down: |\n| [src/transformers/generation\\_tf\\_utils.py](https://codecov.io/gh/huggingface/transformers/pull/5760/diff?src=pr&el=tree#diff-c3JjL3RyYW5zZm9ybWVycy9nZW5lcmF0aW9uX3RmX3V0aWxzLnB5) | `85.96% <0.00%> (-0.26%)` | :arrow_down: |\n| [src/transformers/modeling\\_tf\\_roberta.py](https://codecov.io/gh/huggingface/transformers/pull/5760/diff?src=pr&el=tree#diff-c3JjL3RyYW5zZm9ybWVycy9tb2RlbGluZ190Zl9yb2JlcnRhLnB5) | `93.36% <0.00%> (+49.37%)` | :arrow_up: |\n| [src/transformers/modeling\\_tf\\_mobilebert.py](https://codecov.io/gh/huggingface/transformers/pull/5760/diff?src=pr&el=tree#diff-c3JjL3RyYW5zZm9ybWVycy9tb2RlbGluZ190Zl9tb2JpbGViZXJ0LnB5) | `96.77% <0.00%> (+73.38%)` | :arrow_up: |\n\n------\n\n[Continue to review full report at Codecov](https://codecov.io/gh/huggingface/transformers/pull/5760?src=pr&el=continue).\n> **Legend** - [Click here to learn more](https://docs.codecov.io/docs/codecov-delta)\n> `Δ = absolute <relative> (impact)`, `ø = not affected`, `? = missing data`\n> Powered by [Codecov](https://codecov.io/gh/huggingface/transformers/pull/5760?src=pr&el=footer). Last update [8ab565a...afcf86a](https://codecov.io/gh/huggingface/transformers/pull/5760?src=pr&el=lastupdated). Read the [comment docs](https://docs.codecov.io/docs/pull-request-comments).\n", "> Would it be possible to have one model, e.g. `bart-large-mnli`, linked to the zero shot pipeline and have the others remain linked to text classification?\r\n\r\nYes we can override manually", "So I think I found a possible bug in the calculation of the probabilities. Consider this example:\r\n```python\r\nsequences = [\"Who are you voting for in 2020?\"]\r\ncandidate_labels = [\"politics\", \"public health\", \"economics\", \"elections\"]\r\n\r\nclassifier(sequences[0], candidate_labels)\r\n```\r\nwhich gives the output:\r\n```\r\n{'labels': ['politics', 'elections', 'economics', 'public health'],\r\n 'scores': [0.5225354433059692,\r\n 0.4626988470554352,\r\n 0.007836099714040756,\r\n 0.006929598283022642],\r\n 'sequence': 'Who are you voting for in 2020?'}\r\n```\r\n\r\nI was able to replicate the probability calculations by doing the following:\r\n```python\r\nfrom transformers import AutoTokenizer, AutoModelForSequenceClassification\r\n\r\ntokenizer = AutoTokenizer.from_pretrained(\"facebook/bart-large-mnli\")\r\nmodel = AutoModelForSequenceClassification.from_pretrained(\"facebook/bart-large-mnli\")\r\n\r\nsequences = [\"Who are you voting for in 2020?\"]\r\ncandidate_labels = [\"politics\", \"public health\", \"economics\", \"elections\"]\r\ntemplate = \"This example is {}.\"\r\nx = tokenizer(sequences*len(candidate_labels), [template.format(label) for label in candidate_labels], return_tensors=\"pt\", padding=True)\r\nwith torch.no_grad():\r\n y = model(**x)[0]\r\n probs = F.softmax(y[:,-1], dim=0) # Think this is wrong\r\n\r\nprint(probs)\r\n# tensor([0.5225, 0.0069, 0.0078, 0.4627])\r\n```\r\n\r\nSo I think the probability calculation is wrong. `y` is of shape `(4, 3)` and the softmax should be over the 3 not the the 4 dimensions, since it is over contradiction, neutral, entailment. So what I am suggesting is that the probability calculation should be the following instead:\r\n```python\r\nprobs2 = F.softmax(y, dim=-1)\r\nprobs2 = probs2[:,-1] / sum(probs2[:,-1])\r\nprint(probs2)\r\n# tensor([0.4977, 0.0016, 0.0049, 0.4958])\r\n```\r\n\r\nIt's especially important to do it this way since there is no guarantee that the exponential of `y` will sum to one since they are logits.\r\n", "Hey @sachinruk, thanks for the comment. This isn't a bug though, it's just the way we've chosen to use the outputs. When `multi_class=False`, we ignore the contradiction and neutral logits and just do a softmax over the entailment logits. This does guarantee you will sum to 1. Your snippet is an alternative way of interpreting the model outputs to get probabilities that sum to 1, but I don't see a reason to think it is more correct.", "I have included this in my local Jupyter notebook:\r\n\r\n!pip install git+https://github.com/huggingface/transformers.git --user\r\n\r\nnlp = pipeline(\"zero-shot-classification\") gives the following error.\r\nKeyError: \"Unknown task zero-shot-classification, available tasks are ['feature-extraction', 'sentiment-analysis', 'ner', 'question-answering', 'fill-mask', 'summarization', 'translation_en_to_fr', 'translation_en_to_de', 'translation_en_to_ro', 'text-generation']\"\r\n\r\nWhat could be the missing steps? Thanks.", "@SophiaLeigh you probably just had another version of transformers installed already so the pip install didn’t do anything. Try just passing`--upgrade` with the pip install.", "EDIT: \r\n\r\n**pip install transformers** delivers a version of pipelines.py that is not the one found in the current master\r\n\r\n**pip install pip install git+https://github.com/huggingface/transformers** delivers the correct version obviously. \r\n\r\nI dont know anything about the inner workings of pip but I have the current version _pip 20.2.2_ installed.\r\n\r\n\r\nSame issue as @SophiaLeigh with \r\n\r\n_transformers 3.0.2_ \r\n\r\n---------------------------------------------------------------------------\r\n```\r\nKeyError Traceback (most recent call last)\r\n<ipython-input-6-1f0825594ce1> in <module>\r\n----> 1 classifier = pipeline(\"zero-shot-classification\")\r\n\r\n~/.conda/envs/zero-shot/lib/python3.7/site-packages/transformers/pipelines.py in pipeline(task, model, config, tokenizer, framework, **kwargs)\r\n 1819 # Retrieve the task\r\n 1820 if task not in SUPPORTED_TASKS:\r\n-> 1821 raise KeyError(\"Unknown task {}, available tasks are {}\".format(task, list(SUPPORTED_TASKS.keys())))\r\n 1822 \r\n 1823 framework = framework or get_framework(model)\r\n\r\nKeyError: \"Unknown task zero-shot-classification, available tasks are ['feature-extraction', 'sentiment-analysis', 'ner', 'question-answering', 'fill-mask', 'summarization', 'translation_en_to_fr', 'translation_en_to_de', 'translation_en_to_ro', 'text-generation']\"\r\n \r\n```\r\n\r\n\r\n\r\n\r\nAlso huggingface shows either outdated or wrong info here: \r\n\r\nhttps://huggingface.co/transformers/main_classes/pipelines.html?highlight=pipelines\r\n\r\nand here:\r\n\r\nhttps://huggingface.co/transformers/_modules/transformers/pipelines.html#pipeline\r\n\r\n`Args:\r\n task (:obj:`str`):\r\n The task defining which pipeline will be returned. Currently accepted tasks are:\r\n\r\n - \"feature-extraction\": will return a :class:`~transformers.FeatureExtractionPipeline`\r\n - \"sentiment-analysis\": will return a :class:`~transformers.TextClassificationPipeline`\r\n - \"ner\": will return a :class:`~transformers.TokenClassificationPipeline`\r\n - \"question-answering\": will return a :class:`~transformers.QuestionAnsweringPipeline`\r\n - \"fill-mask\": will return a :class:`~transformers.FillMaskPipeline`\r\n - \"summarization\": will return a :class:`~transformers.SummarizationPipeline`\r\n - \"translation_xx_to_yy\": will return a :class:`~transformers.TranslationPipeline`\r\n - \"text-generation\": will return a :class:`~transformers.TextGenerationPipeline``\r\n\r\nWhile I can clearly see it`s here :)\r\n\r\nhttps://github.com/huggingface/transformers/blob/master/src/transformers/pipelines.py\r\n\r\nI installed via pip install transformers and also via pip install git+https/thisrepository and ~~both versions~~ have the correct pipelines.py file that has all the implementations for zero-shot- Now I am really confused. ", "the error persists though I have included --upgrade \r\n\r\n`pip install git+https://github.com/huggingface/transformers --upgrade\r\n!pip install git+https://github.com/huggingface/transformers --upgrade\r\nCollecting git+https://github.com/huggingface/transformers\r\n Cloning https://github.com/huggingface/transformers to c:\\users\\sophia\\appdata\\local\\temp\\pip-req-build-6_7vqcz3\r\nRequirement already satisfied, skipping upgrade: numpy in c:\\users\\sophia\\anaconda3\\lib\\site-packages (from transformers==3.0.2) (1.19.1)\r\nRequirement already satisfied, skipping upgrade: tokenizers==0.8.1.rc2 in c:\\users\\sophia\\anaconda3\\lib\\site-packages (from transformers==3.0.2) (0.8.1rc2)\r\nRequirement already satisfied, skipping upgrade: packaging in c:\\users\\sophia\\anaconda3\\lib\\site-packages (from transformers==3.0.2) (20.4)\r\nRequirement already satisfied, skipping upgrade: filelock in c:\\users\\sophia\\anaconda3\\lib\\site-packages (from transformers==3.0.2) (3.0.12)\r\nRequirement already satisfied, skipping upgrade: requests in c:\\users\\sophia\\anaconda3\\lib\\site-packages (from transformers==3.0.2) (2.24.0)\r\nRequirement already satisfied, skipping upgrade: tqdm>=4.27 in c:\\users\\sophia\\anaconda3\\lib\\site-packages (from transformers==3.0.2) (4.48.2)\r\nRequirement already satisfied, skipping upgrade: regex!=2019.12.17 in c:\\users\\sophia\\anaconda3\\lib\\site-packages (from transformers==3.0.2) (2020.7.14)\r\nRequirement already satisfied, skipping upgrade: sentencepiece!=0.1.92 in c:\\users\\sophia\\anaconda3\\lib\\site-packages (from transformers==3.0.2) (0.1.91)\r\nRequirement already satisfied, skipping upgrade: sacremoses in c:\\users\\sophia\\anaconda3\\lib\\site-packages (from transformers==3.0.2) (0.0.43)\r\nRequirement already satisfied, skipping upgrade: six in c:\\users\\sophia\\anaconda3\\lib\\site-packages (from packaging->transformers==3.0.2) (1.15.0)\r\nRequirement already satisfied, skipping upgrade: pyparsing>=2.0.2 in c:\\users\\sophia\\anaconda3\\lib\\site-packages (from packaging->transformers==3.0.2) (2.4.7)\r\nRequirement already satisfied, skipping upgrade: urllib3!=1.25.0,!=1.25.1,<1.26,>=1.21.1 in c:\\users\\sophia\\anaconda3\\lib\\site-packages (from requests->transformers==3.0.2) (1.25.9)\r\nRequirement already satisfied, skipping upgrade: certifi>=2017.4.17 in c:\\users\\sophia\\anaconda3\\lib\\site-packages (from requests->transformers==3.0.2) (2020.6.20)\r\nRequirement already satisfied, skipping upgrade: chardet<4,>=3.0.2 in c:\\users\\sophia\\anaconda3\\lib\\site-packages (from requests->transformers==3.0.2) (3.0.4)\r\nRequirement already satisfied, skipping upgrade: idna<3,>=2.5 in c:\\users\\sophia\\anaconda3\\lib\\site-packages (from requests->transformers==3.0.2) (2.10)\r\nRequirement already satisfied, skipping upgrade: joblib in c:\\users\\sophia\\anaconda3\\lib\\site-packages (from sacremoses->transformers==3.0.2) (0.16.0)\r\nRequirement already satisfied, skipping upgrade: click in c:\\users\\sophia\\anaconda3\\lib\\site-packages (from sacremoses->transformers==3.0.2) (7.1.2)\r\nBuilding wheels for collected packages: transformers\r\n Building wheel for transformers (setup.py): started\r\n Building wheel for transformers (setup.py): finished with status 'done'\r\n Created wheel for transformers: filename=transformers-3.0.2-py3-none-any.whl size=873793 sha256=c363c4ea6e4ec438b88c6fafb162915d24a2c149dd4210111713e78496940af2\r\n Stored in directory: C:\\Users\\sophia\\AppData\\Local\\Temp\\pip-ephem-wheel-cache-ck6yr140\\wheels\\35\\2e\\a7\\d819e3310040329f0f47e57c9e3e7a7338aa5e74c49acfe522\r\nSuccessfully built transformers\r\nInstalling collected packages: transformers\r\n Attempting uninstall: transformers\r\n Found existing installation: transformers 3.0.2\r\n Uninstalling transformers-3.0.2:\r\n Successfully uninstalled transformers-3.0.2\r\nSuccessfully installed transformers-3.0.2`", "@SophiaLei did you restart your kernel after upgrading?\r\n@Tabernakel In the top left corner of the docs, click on `v3.0.2` and switch to `master`.\r\n\r\nFeel free to open a topic on https://discuss.huggingface.co with any other questions.", "It works now. Thank you.", "So the online demo has two different models MNLI and MLNI + Yahoo Answers. I know the second one is Bart with a classification head trained on MNLI and then further fine-tuned on Yahoo Answers topic classification. Is there a specific scenario where MNLI + yahoo answers outperform just MNLI in the zero-shot classification task? ", "@avinregmi For practical usage the base MNLI model is typically going to do better. The Yahoo Answers model will do better on Yahoo Answers, which seems unhelpful unless you recognize that was only fine-tuned on 5 labels out of the 10 in the corpus. So if you have a big labeled dataset but only covering a subset of the labels you want to be able classify into, fine-tuning an MNLI model as I did with Yahoo Answers will likely boost your performance. Otherwise, stick with the base MNLI model.", "Is there a way to persist Zero shot classification pipeline and use it for deploying in production?\r\n\r\nThanks!", "> Is there a way to persist Zero shot classification pipeline and use it for deploying in production?\r\n\r\n@mariyamiteva You have two options:\r\n\r\n1. Use the pipeline directly via our inference API and let us (Hugging Face engineers) take care of all the production serving for you. Check out [the documentation for our inference API](https://api-inference.huggingface.co/docs/python/html/index.html) and reach out to [email protected] to discuss further if you're interested. cc @jeffboudier \r\n\r\n2. Use [this distillation script](https://github.com/huggingface/transformers/tree/master/examples/research_projects/zero-shot-distillation) I wrote with some unlabeled data to distill the zero-shot pipeline to a more efficient sequence classification model, which can then be shared and deployed like any model.", "Thanks @joeddav - and @mariyamiteva you can do both: 2) Distill the model to efficient sentence classification and then 1) upload it as a private model on Hugging Face to serve it via our hosted inference API.", "Thanks @joeddav and @jeffboudier for the prompt feedback!\r\n\r\nI am currently trying to run `distill_classifier.py`. `roberta-large-mnli` has been used as a teacher:\r\n\r\n python distill_classifier.py\r\n\r\n --data_file ..\\model\\output\\unlabeled.txt \r\n --class_names_file ..\\model\\output\\class_names.txt\r\n --teacher_name_or_path roberta-large-mnli\r\n --multi_label 1\r\n --output_dir ..\\model\\output\\distilled\r\n\r\nMy `unlabeled.txt` has a single line text, e.g. \r\n\r\n\r\n\r\n\r\nMy `class_names.txt` takes the following form:\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\nHere is the snippet of the error I get:\r\n\r\n File \"distill_classifier.py\", line 338, in <module>\r\n main()\r\n File \"distill_classifier.py\", line 328, in main\r\n trainer.train()\r\n File \"C:\\Users\\Maria\\anaconda3\\lib\\site-packages\\transformers\\trainer.py\", line 1222, in train\r\n tr_loss += self.training_step(model, inputs)\r\n File \"C:\\Users\\Maria\\anaconda3\\lib\\site-packages\\transformers\\trainer.py\", line 1617, in training_step\r\n loss = self.compute_loss(model, inputs)\r\n File \"distill_classifier.py\", line 119, in compute_loss\r\n target_p = inputs[\"labels\"]\r\n File \"C:\\Users\\Maria\\anaconda3\\lib\\site-packages\\transformers\\tokenization_utils_base.py\", line 230, in __getitem__\r\n return self.data[item]\r\n KeyError: 'labels'\r\n\r\nUnfortunately, I was not able to deal with the error above. Could you please help? Thanks!", "@mariyamiteva I'm not positive this is the issue, but you might need to end your `unlabeled.txt` with a newline. Also, you don't need the `'` single quotes around your text or class names.", "Yes, it was not the issue. I performed the proposed changes and upgraded the version to 4.7.0.dev0. \r\n\r\nThe error occurred is not the same, but similar:\r\n\r\n Traceback (most recent call last):\r\n File \"distill_classifier.py\", line 338, in <module>\r\n main()\r\n File \"distill_classifier.py\", line 328, in main\r\n trainer.train()\r\n File \"C:\\Users\\Maria\\anaconda3\\lib\\site-packages\\transformers\\trainer.py\", line 1261, in train\r\n tr_loss += self.training_step(model, inputs)\r\n File \"C:\\Users\\Maria\\anaconda3\\lib\\site-packages\\transformers\\trainer.py\", line 1734, in training_step\r\n loss = self.compute_loss(model, inputs)\r\n File \"distill_classifier.py\", line 119, in compute_loss\r\n target_p = inputs[\"labels\"]\r\n KeyError: 'labels'", "@mariyamiteva I'm not sure this is this issue, but do you have tokenizers installed? Try `pip install tokenizers` and let me know what that gives you.", "@joeddav Could the following transformer model used for zero-shot classification be optimized in terms of model inference through ONNX - ‘joeddav/xlm-roberta-large-xnli’?\r\n\r\nThanks!\r\nM.", "@joeddav What if anyone wants to fine-tune the zero-shot classifier for a specific domain dataset. Is there any code or GitHub repo that may help us to train the zero-shot classifier?" ]

[ { "id": 1838412367, "node_id": "MDU6TGFiZWwxODM4NDEyMzY3", "url": "https://api.github.com/repos/huggingface/transformers/labels/model%20card", "name": "model card", "color": "92d5f4", "default": false, "description": "Related to pretrained model cards" } ]

[]

[ "The file paths aren't correct. (we have a \"legacy\" format for non-namespaced models – model pages link to the correct paths)\r\n\r\nIn fact if you look at the model pages, e.g. https://huggingface.co/t5-base – they already have (non-textual) model cards. Can you update those? Thanks=)", "Up, @sshleifer ", "# [Codecov](https://codecov.io/gh/huggingface/transformers/pull/5759?src=pr&el=h1) Report\n> Merging [#5759](https://codecov.io/gh/huggingface/transformers/pull/5759?src=pr&el=desc) into [master](https://codecov.io/gh/huggingface/transformers/commit/ae67b2439fb15954bfd8f0fdf521cf1a650bafb9&el=desc) will **increase** coverage by `0.17%`.\n> The diff coverage is `n/a`.\n\n[](https://codecov.io/gh/huggingface/transformers/pull/5759?src=pr&el=tree)\n\n```diff\n@@ Coverage Diff @@\n## master #5759 +/- ##\n==========================================\n+ Coverage 78.51% 78.69% +0.17% \n==========================================\n Files 146 146 \n Lines 26214 26214 \n==========================================\n+ Hits 20581 20628 +47 \n+ Misses 5633 5586 -47 \n```\n\n\n| [Impacted Files](https://codecov.io/gh/huggingface/transformers/pull/5759?src=pr&el=tree) | Coverage Δ | |\n|---|---|---|\n| [src/transformers/modeling\\_tf\\_distilbert.py](https://codecov.io/gh/huggingface/transformers/pull/5759/diff?src=pr&el=tree#diff-c3JjL3RyYW5zZm9ybWVycy9tb2RlbGluZ190Zl9kaXN0aWxiZXJ0LnB5) | `64.90% <0.00%> (-33.90%)` | :arrow_down: |\n| [src/transformers/file\\_utils.py](https://codecov.io/gh/huggingface/transformers/pull/5759/diff?src=pr&el=tree#diff-c3JjL3RyYW5zZm9ybWVycy9maWxlX3V0aWxzLnB5) | `81.19% <0.00%> (ø)` | |\n| [src/transformers/modeling\\_dpr.py](https://codecov.io/gh/huggingface/transformers/pull/5759/diff?src=pr&el=tree#diff-c3JjL3RyYW5zZm9ybWVycy9tb2RlbGluZ19kcHIucHk=) | `97.83% <0.00%> (ø)` | |\n| [src/transformers/modeling\\_gpt2.py](https://codecov.io/gh/huggingface/transformers/pull/5759/diff?src=pr&el=tree#diff-c3JjL3RyYW5zZm9ybWVycy9tb2RlbGluZ19ncHQyLnB5) | `85.88% <0.00%> (ø)` | |\n| [src/transformers/modeling\\_albert.py](https://codecov.io/gh/huggingface/transformers/pull/5759/diff?src=pr&el=tree#diff-c3JjL3RyYW5zZm9ybWVycy9tb2RlbGluZ19hbGJlcnQucHk=) | `82.04% <0.00%> (ø)` | |\n| [src/transformers/modeling\\_openai.py](https://codecov.io/gh/huggingface/transformers/pull/5759/diff?src=pr&el=tree#diff-c3JjL3RyYW5zZm9ybWVycy9tb2RlbGluZ19vcGVuYWkucHk=) | `82.31% <0.00%> (ø)` | |\n| [src/transformers/modeling\\_electra.py](https://codecov.io/gh/huggingface/transformers/pull/5759/diff?src=pr&el=tree#diff-c3JjL3RyYW5zZm9ybWVycy9tb2RlbGluZ19lbGVjdHJhLnB5) | `81.55% <0.00%> (ø)` | |\n| [src/transformers/modeling\\_longformer.py](https://codecov.io/gh/huggingface/transformers/pull/5759/diff?src=pr&el=tree#diff-c3JjL3RyYW5zZm9ybWVycy9tb2RlbGluZ19sb25nZm9ybWVyLnB5) | `89.21% <0.00%> (ø)` | |\n| [src/transformers/modeling\\_mobilebert.py](https://codecov.io/gh/huggingface/transformers/pull/5759/diff?src=pr&el=tree#diff-c3JjL3RyYW5zZm9ybWVycy9tb2RlbGluZ19tb2JpbGViZXJ0LnB5) | `89.45% <0.00%> (ø)` | |\n| [src/transformers/tokenization\\_utils\\_base.py](https://codecov.io/gh/huggingface/transformers/pull/5759/diff?src=pr&el=tree#diff-c3JjL3RyYW5zZm9ybWVycy90b2tlbml6YXRpb25fdXRpbHNfYmFzZS5weQ==) | `93.60% <0.00%> (ø)` | |\n| ... and [2 more](https://codecov.io/gh/huggingface/transformers/pull/5759/diff?src=pr&el=tree-more) | |\n\n------\n\n[Continue to review full report at Codecov](https://codecov.io/gh/huggingface/transformers/pull/5759?src=pr&el=continue).\n> **Legend** - [Click here to learn more](https://docs.codecov.io/docs/codecov-delta)\n> `Δ = absolute <relative> (impact)`, `ø = not affected`, `? = missing data`\n> Powered by [Codecov](https://codecov.io/gh/huggingface/transformers/pull/5759?src=pr&el=footer). Last update [ae67b24...583295f](https://codecov.io/gh/huggingface/transformers/pull/5759?src=pr&el=lastupdated). Read the [comment docs](https://docs.codecov.io/docs/pull-request-comments).\n", "CircleCI jobs say they succeeded if you click through. Unclear why they are yellow on this page." ]

[ { "id": 1838412367, "node_id": "MDU6TGFiZWwxODM4NDEyMzY3", "url": "https://api.github.com/repos/huggingface/transformers/labels/model%20card", "name": "model card", "color": "92d5f4", "default": false, "description": "Related to pretrained model cards" } ]

[]

[ "# [Codecov](https://codecov.io/gh/huggingface/transformers/pull/5758?src=pr&el=h1) Report\n> Merging [#5758](https://codecov.io/gh/huggingface/transformers/pull/5758?src=pr&el=desc) into [master](https://codecov.io/gh/huggingface/transformers/commit/8ab565a4be5a7fd96b19ef88d474037ef31f27e5&el=desc) will **not change** coverage.\n> The diff coverage is `n/a`.\n\n[](https://codecov.io/gh/huggingface/transformers/pull/5758?src=pr&el=tree)\n\n```diff\n@@ Coverage Diff @@\n## master #5758 +/- ##\n=======================================\n Coverage 77.32% 77.32% \n=======================================\n Files 146 146 \n Lines 26047 26047 \n=======================================\n Hits 20141 20141 \n Misses 5906 5906 \n```\n\n\n\n------\n\n[Continue to review full report at Codecov](https://codecov.io/gh/huggingface/transformers/pull/5758?src=pr&el=continue).\n> **Legend** - [Click here to learn more](https://docs.codecov.io/docs/codecov-delta)\n> `Δ = absolute <relative> (impact)`, `ø = not affected`, `? = missing data`\n> Powered by [Codecov](https://codecov.io/gh/huggingface/transformers/pull/5758?src=pr&el=footer). Last update [8ab565a...a67f412](https://codecov.io/gh/huggingface/transformers/pull/5758?src=pr&el=lastupdated). Read the [comment docs](https://docs.codecov.io/docs/pull-request-comments).\n" ]

[ { "id": 1314768611, "node_id": "MDU6TGFiZWwxMzE0NzY4NjEx", "url": "https://api.github.com/repos/huggingface/transformers/labels/wontfix", "name": "wontfix", "color": "ffffff", "default": true, "description": null } ]

[ { "login": "yjernite", "id": 10469459, "node_id": "MDQ6VXNlcjEwNDY5NDU5", "avatar_url": "https://avatars.githubusercontent.com/u/10469459?v=4", "gravatar_id": "", "url": "https://api.github.com/users/yjernite", "html_url": "https://github.com/yjernite", "followers_url": "https://api.github.com/users/yjernite/followers", "following_url": "https://api.github.com/users/yjernite/following{/other_user}", "gists_url": "https://api.github.com/users/yjernite/gists{/gist_id}", "starred_url": "https://api.github.com/users/yjernite/starred{/owner}{/repo}", "subscriptions_url": "https://api.github.com/users/yjernite/subscriptions", "organizations_url": "https://api.github.com/users/yjernite/orgs", "repos_url": "https://api.github.com/users/yjernite/repos", "events_url": "https://api.github.com/users/yjernite/events{/privacy}", "received_events_url": "https://api.github.com/users/yjernite/received_events", "type": "User", "site_admin": false } ]

[ "This issue has been automatically marked as stale because it has not had recent activity. It will be closed if no further activity occurs. Thank you for your contributions.\n" ]

[ { "id": 1314768611, "node_id": "MDU6TGFiZWwxMzE0NzY4NjEx", "url": "https://api.github.com/repos/huggingface/transformers/labels/wontfix", "name": "wontfix", "color": "ffffff", "default": true, "description": null } ]

[ { "login": "joeddav", "id": 9353833, "node_id": "MDQ6VXNlcjkzNTM4MzM=", "avatar_url": "https://avatars.githubusercontent.com/u/9353833?v=4", "gravatar_id": "", "url": "https://api.github.com/users/joeddav", "html_url": "https://github.com/joeddav", "followers_url": "https://api.github.com/users/joeddav/followers", "following_url": "https://api.github.com/users/joeddav/following{/other_user}", "gists_url": "https://api.github.com/users/joeddav/gists{/gist_id}", "starred_url": "https://api.github.com/users/joeddav/starred{/owner}{/repo}", "subscriptions_url": "https://api.github.com/users/joeddav/subscriptions", "organizations_url": "https://api.github.com/users/joeddav/orgs", "repos_url": "https://api.github.com/users/joeddav/repos", "events_url": "https://api.github.com/users/joeddav/events{/privacy}", "received_events_url": "https://api.github.com/users/joeddav/received_events", "type": "User", "site_admin": false } ]

[ "This issue has been automatically marked as stale because it has not had recent activity. It will be closed if no further activity occurs. Thank you for your contributions.\n" ]

[ { "id": 1314768611, "node_id": "MDU6TGFiZWwxMzE0NzY4NjEx", "url": "https://api.github.com/repos/huggingface/transformers/labels/wontfix", "name": "wontfix", "color": "ffffff", "default": true, "description": null } ]

[ { "login": "sshleifer", "id": 6045025, "node_id": "MDQ6VXNlcjYwNDUwMjU=", "avatar_url": "https://avatars.githubusercontent.com/u/6045025?v=4", "gravatar_id": "", "url": "https://api.github.com/users/sshleifer", "html_url": "https://github.com/sshleifer", "followers_url": "https://api.github.com/users/sshleifer/followers", "following_url": "https://api.github.com/users/sshleifer/following{/other_user}", "gists_url": "https://api.github.com/users/sshleifer/gists{/gist_id}", "starred_url": "https://api.github.com/users/sshleifer/starred{/owner}{/repo}", "subscriptions_url": "https://api.github.com/users/sshleifer/subscriptions", "organizations_url": "https://api.github.com/users/sshleifer/orgs", "repos_url": "https://api.github.com/users/sshleifer/repos", "events_url": "https://api.github.com/users/sshleifer/events{/privacy}", "received_events_url": "https://api.github.com/users/sshleifer/received_events", "type": "User", "site_admin": false } ]

[ "Thanks for flagging. We are still trying to get to the bottom of which special tokens to use for mbart-large-cc25. See https://github.com/pytorch/fairseq/issues/2258 .", "This might not be the best solution but after experimenting with the tokenizer special tokens a bit, it seems like the model is insensitive to the first input_id and lang_code used on the encoder side.\r\nSo after these modifications:\r\n\r\n```\r\n def set_src_lang_special_tokens(self, src_lang) -> None:\r\n \"\"\"Reset the special tokens to the source lang setting. No prefix and suffix=[eos, cur_lang_code].\"\"\"\r\n self.cur_lang_code = self.lang_code_to_id[src_lang]\r\n self.prefix_tokens = [self.bos_token_id]\r\n self.suffix_tokens = [self.eos_token_id]\r\n\r\n def set_tgt_lang_special_tokens(self, lang: str) -> None:\r\n \"\"\"Reset the special tokens to the target language setting. Prefix [tgt_lang_code], suffix =[eos].\"\"\"\r\n self.cur_lang_code = self.lang_code_to_id[lang]\r\n self.prefix_tokens = [self.cur_lang_code]\r\n self.suffix_tokens = [self.eos_token_id]\r\n```\r\n\r\nin tokenization_bart.py, the model seems to be doing the right thing and generating correct English output:\r\n```\r\nsrc_sent: UN Chief Says There Is No Military Solution in Syria\r\nsrc_ids: {'input_ids': tensor([[ 0, 8274, 127873, 25916, 7, 8622, 2071, 438, 67485,\r\n 53, 187895, 23, 51712, 2]]), 'attention_mask': tensor([[1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1]])}\r\noutput_ids: tensor([[250004, 0, 8274, 127873, 25916, 7, 8622, 2071, 438,\r\n 67485, 53, 187895, 23, 51712]])\r\noutput: UN Chief Says There Is No Military Solution in Syria\r\n```\r\n\r\nAlthough this is not how things are described in MBART paper so the core issue remains. \r\n\r\nAlso, this code segment in modeling_bart.py :\r\n\r\n```\r\n def adjust_logits_during_generation(self, logits, cur_len, max_length, **kwargs):\r\n if cur_len == 1:\r\n self._force_token_ids_generation(logits, self.config.bos_token_id)\r\n```\r\nis the culprit for generating 0 (corresponding to bos_token) at every first decoding step.\r\nIt might need to be changed for MBART model.\r\n\r\n", "I find that the input does need to start with `<bos>`, and the decoder should be seeded with `<lang_code> <bos>`. With this setup, I am able to recover the input sequence during decoding. Like @Mehrad0711 I find that the input `lang_code` does not make a significant difference. ", "@tomhosking I replicated what you wrote. Definitely needs `<s>`(bos) at the beginning of the input string to fix the off by one error. You can see the tests in this PR #6524 .\r\n\r\nI'm still having trouble squaring this/trying to find a unified fix to accomodate the behavior that mbart-large-en-ro seems to want, as shown in #6156 .\r\n\r\nMaybe the simplest change is just to add `<s>` to the start of the encoder side string?", "Interestingly, `prepend_bos` is set to false by default in the fairseq mbart finetuning docs.\r\nI set a breakpoint during finetuning and there is no BOS to be found: here is [how batches look](https://gist.github.com/sshleifer/cba08bc2109361a74ac3760a7e30e4f4)", "This issue has been automatically marked as stale because it has not had recent activity. It will be closed if no further activity occurs. Thank you for your contributions.\n", "Hi,\r\nReviving this issue as it still persists in the recent version of transformers. The solution I proposed in Aug 2 comment seems to be working after a modification on prefix source token borrowed from `tokenization_mbart50.py`.\r\nProposed solution:\r\n```\r\n def set_src_lang_special_tokens(self, src_lang) -> None:\r\n \"\"\"Reset the special tokens to the source lang setting. Prefix=[src_lang_code] and suffix=[eos].\"\"\"\r\n self.cur_lang_code = self.lang_code_to_id[src_lang]\r\n self.prefix_tokens = [self.cur_lang_code]\r\n self.suffix_tokens = [self.eos_token_id]\r\n\r\n def set_tgt_lang_special_tokens(self, lang: str) -> None:\r\n \"\"\"Reset the special tokens to the target language setting. Prefix=[tgt_lang_code] and suffix=[eos].\"\"\"\r\n self.cur_lang_code = self.lang_code_to_id[lang]\r\n self.prefix_tokens = [self.cur_lang_code]\r\n self.suffix_tokens = [self.eos_token_id]\r\n```\r\nThis change will fix `mbart-large-cc25`'s output text while leaving `mbart-large-en-ro`'s untouched.\r\n\r\nCode to reproduce:\r\n```\r\nfrom transformers import MBartTokenizer, MBartForConditionalGeneration\r\n\r\ntokenizer = MBartTokenizer.from_pretrained('facebook/mbart-large-cc25')\r\nmodel = MBartForConditionalGeneration.from_pretrained('facebook/mbart-large-cc25')\r\n\r\nsrc_sent = \"UN Chief Says There Is No Military Solution in Syria\"\r\n\r\nbatch = tokenizer.prepare_seq2seq_batch(src_texts=[src_sent], src_lang=\"en_XX\", return_tensors=\"pt\")\r\n\r\noutput_ids = model.generate(**batch, decoder_start_token_id=tokenizer.lang_code_to_id[\"en_XX\"])\r\n\r\noutput = tokenizer.batch_decode(output_ids, skip_special_tokens=True)[0]\r\n\r\nprint('src_sent: ', src_sent)\r\nprint('src_ids: ', batch[\"input_ids\"])\r\nprint('output_ids: ', output_ids)\r\nprint('output: ', output)\r\n\r\n```\r\n\r\nstdout (before change):\r\n```\r\nsrc_sent: UN Chief Says There Is No Military Solution in Syria\r\nsrc_ids: tensor([[ 8274, 127873, 25916, 7, 8622, 2071, 438, 67485, 53,\r\n 187895, 23, 51712, 2, 250004]])\r\noutput_ids: tensor([[250004, 0, 127873, 25916, 7, 8622, 2071, 438, 67485,\r\n 53, 187895, 23, 51712, 2]])\r\noutput: Chief Says There Is No Military Solution in Syria\r\n```\r\n\r\nstdout (after change):\r\n```\r\nsrc_sent: UN Chief Says There Is No Military Solution in Syria\r\nsrc_ids: tensor([[250004, 8274, 127873, 25916, 7, 8622, 2071, 438, 67485,\r\n 53, 187895, 23, 51712, 2]])\r\noutput_ids: tensor([[250004, 0, 8274, 127873, 25916, 7, 8622, 2071, 438,\r\n 67485, 53, 187895, 23, 51712, 2]])\r\noutput: UN Chief Says There Is No Military Solution in Syria\r\n\r\n```\r\n\r\nPotential reviewers: @patrickvonplaten, @patil-suraj, @sgugger \r\n", "Hi @sgugger, @patrickvonplaten, @patil-suraj,\r\nit would be great if you could provide your feedback on this. I would be happy to provide more context if needed.", "Hi @Mehrad0711 \r\n\r\nThank you for reporting this. I'll go through the original model code in fairseq to see how they are handling the prefix tokens and get back here.", "This issue has been automatically marked as stale because it has not had recent activity. If you think this still needs to be addressed please comment on this thread.\n\nPlease note that issues that do not follow the [contributing guidelines](https://github.com/huggingface/transformers/blob/master/CONTRIBUTING.md) are likely to be ignored.", "Hi @patil-suraj,\r\nWas wondering if you had the chance to take a look at this. Thanks.", "Hi @Mehrad0711 \r\n\r\nYou are right, all mBART models actually use the language code as the prefix token and `<eos>` as the suffix token.\r\nBut unfortunately, we can't really change it now, because this will be backward incompatible with the other models trained using the existing format.\r\n\r\nAlso, this doesn't really make that much difference if you want to fine-tune the model. As long as a consistent format is used for fine-tuning and then for inference then it should work. However, it would change the output for the pre-trained models (as you reported). But as `mbart-large-cc25` is just a pre-trained model and should be fine-tuned to use it for the downstream tasks, this doesn't seem like a big issue.\r\n", "Hi @patil-suraj!\r\nThanks for your reply. I understand the concern regarding backward incompatibility of this change.\r\nI was using `mbart-large-cc25` without fine-tuning for text denoising; that's how the problem popped up. Given newer mbart models are now available on huggingface, I'll switch to them." ]